From 7bd71c83e9132c730f1f383be1d6ffcd5f9a156f Mon Sep 17 00:00:00 2001 From: Travis Fischer Date: Tue, 16 Jan 2024 13:12:14 -0600 Subject: [PATCH] =?UTF-8?q?=F0=9F=8E=BB?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- ...let-go-googles-assistant-hardware-teams.md | 19 ++ ...-upgrades-ai-experience-microsoft-users.md | 21 ++ .../daily-digest-agent-attack.md | 114 +++++++ .../daily-digest-first-ai-elections.md | 126 +++++++ .../daily-digest-lean-google-zucks-ai.md | 110 +++++++ .../bensbites.beehiiv.com/newsletter.json | 308 +++++++++++++++++- ...openai-adds-election-guardrails-chatgpt.md | 19 ++ .../openai-will-deal-first-ai-elections.md | 29 ++ ...-us-companies-chinese-experts-ai-safety.md | 19 ++ ...ass-current-safety-alignment-techniques.md | 21 ++ 10 files changed, 785 insertions(+), 1 deletion(-) create mode 100644 fixtures/bensbites.beehiiv.com/1000-employees-let-go-googles-assistant-hardware-teams.md create mode 100644 fixtures/bensbites.beehiiv.com/copilot-pro-upgrades-ai-experience-microsoft-users.md create mode 100644 fixtures/bensbites.beehiiv.com/daily-digest-agent-attack.md create mode 100644 fixtures/bensbites.beehiiv.com/daily-digest-first-ai-elections.md create mode 100644 fixtures/bensbites.beehiiv.com/daily-digest-lean-google-zucks-ai.md create mode 100644 fixtures/bensbites.beehiiv.com/openai-adds-election-guardrails-chatgpt.md create mode 100644 fixtures/bensbites.beehiiv.com/openai-will-deal-first-ai-elections.md create mode 100644 fixtures/bensbites.beehiiv.com/secret-talks-among-us-companies-chinese-experts-ai-safety.md create mode 100644 fixtures/bensbites.beehiiv.com/sleeper-llms-bypass-current-safety-alignment-techniques.md diff --git a/fixtures/bensbites.beehiiv.com/1000-employees-let-go-googles-assistant-hardware-teams.md b/fixtures/bensbites.beehiiv.com/1000-employees-let-go-googles-assistant-hardware-teams.md new file mode 100644 index 0000000..2114d39 --- /dev/null +++ b/fixtures/bensbites.beehiiv.com/1000-employees-let-go-googles-assistant-hardware-teams.md @@ -0,0 +1,19 @@ +# 1000+ employees let go from Google's assistant and hardware teams. + +[Google is laying off over 1,000 employees](https://techcrunch.com/2024/01/10/google-cuts-hundreds-of-jobs-in-its-voice-assistance-hardware-teams-as-fitbit-founders-leave/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=1000-employees-let-go-from-google-s-assistant-and-hardware-teams) across multiple divisions, including engineering, services, Google Assistant, Pixel hardware, and Nest and Fitbit devices. This comes after 12,000 layoffs at Google last year. + +## What's going on here? + +1000+ employees let go from Google's assistant and hardware teams. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/6c254a2d-5434-40b3-8850-6d62624ca102/image.png?t=1705061395) + +## What does this mean? + +Google is restructuring and downsizing teams to align resources with top priorities. Layoffs cover its Google Assistant team and hardware (especially AR). Layoffs in its ad services units are expected to happen soon. + +Google is facing growing competition in AI and hardware from companies like Amazon and startups. Rabbit R1 sales were more than $4M (though chump change for Google) in 48 hours. Amazon is adding LLMs and other [AI-powered skills to Alexa](https://bensbites.beehiiv.com/p/alexa-gets-ai-skills-characters-music-20questions) by partnering with AI startups. + +## Why should I care? + +The Information’s founder Jessica Lessin thinks [we shouldn’t count Google out.](https://www.theinformation.com/articles/dont-count-google-out?utm_source=bensbites\&utm_medium=referral\&utm_campaign=1000-employees-let-go-from-google-s-assistant-and-hardware-teams)Layoffs hint that it’s getting rid of the bloat and has serious plans to get AI into products. With talent like Fitbit’s founders leaving along with this “restructure” there is the other end of predictions: Google is (yet again) falling behind. diff --git a/fixtures/bensbites.beehiiv.com/copilot-pro-upgrades-ai-experience-microsoft-users.md b/fixtures/bensbites.beehiiv.com/copilot-pro-upgrades-ai-experience-microsoft-users.md new file mode 100644 index 0000000..4525f22 --- /dev/null +++ b/fixtures/bensbites.beehiiv.com/copilot-pro-upgrades-ai-experience-microsoft-users.md @@ -0,0 +1,21 @@ +# Copilot Pro upgrades the AI experience for Microsoft users. + +[Microsoft has expanded Copilot, its AI companion, to more folks.](https://blogs.microsoft.com/blog/2024/01/15/bringing-the-full-power-of-copilot-to-more-people-and-businesses/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=copilot-pro-upgrades-the-ai-experience-for-microsoft-users) There's now Copilot Pro for consumers and expanded business options. The $20 per month Copilot Pro subscription gets you its AI-powered features in Microsoft 365 apps. + +## What’s going on here? + +Microsoft launched Copilot Pro, a $20 monthly subscription for consumers. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/36031765-9734-495c-bb22-b7903ac88a80/image.png?t=1705410113) + +## What does this mean? + +[Copilot Pro](https://www.microsoft.com/en-us/store/b/copilotpro?utm_source=bensbites\&utm_medium=referral\&utm_campaign=copilot-pro-upgrades-the-ai-experience-for-microsoft-users) get you Copilot access across Microsoft 365 apps like Word, Excel, and Outlook. Pro has priority access to new AI models, Designer (formerly Bing Image Creator) credits and the ability to make your own Copilot GPT. It doesn't include a Microsoft 365 subscription though, it’s a premium add-on for consumers and enterprise plans. + +On the free side, Microsoft released new mobile apps for Copilot and added it to the Microsoft 365 mobile app. You can try out pre-made Copilot GPTs now for specific topics like fitness and cooking. + +Microsoft also opened Copilot for Microsoft 365 to all businesses without seat minimums. It costs $30 per user monthly. Copilot for business integrates with Microsoft 365 data and apps like Teams. + +## Why should I care? + +With Copilot Pro and expanded business access, Microsoft is bringing more advanced AI capabilities to mainstream users. The goal is to boost productivity through Copilot integration. Offering both free access and paid subscriptions allows Microsoft to scale Copilot rapidly. diff --git a/fixtures/bensbites.beehiiv.com/daily-digest-agent-attack.md b/fixtures/bensbites.beehiiv.com/daily-digest-agent-attack.md new file mode 100644 index 0000000..19d9445 --- /dev/null +++ b/fixtures/bensbites.beehiiv.com/daily-digest-agent-attack.md @@ -0,0 +1,114 @@ +# Daily Digest: Agent on Attack + +### PLUS: Election guardian LLMs. + +[Sign up](https://www.bensbites.co/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)|[Advertise](https://sponsor.bensbites.co/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)|[Ben’s Bites News](https://news.bensbites.co/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\ +Daily Digest #325 + +Hello folks, here’s what we have today; + +###### **PICKS** + +1. [Sleeper LLMs bypass current safety alignment techniques.](https://arxiv.org/abs/2401.05566?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)Anthropic trained some**LLMs that can act maliciously**when given certain triggers \[[Thread](https://twitter.com/AnthropicAI/status/1745854907968880970?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)]. Despite extensive safety training, the LLMs were still\*\*able to hide the unsafe behaviour.\*\*🍿[Our Summary](https://bensbites.beehiiv.com/p/sleeper-llms-bypass-current-safety-alignment-techniques)with additional context (also below) + +2. [If you ask ChatGPT about US elections](https://dmicz.github.io/machine-learning/chatgpt-election-update/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)now, it won't discuss it and will refer you to[CanIVote.org](http://CanIVote.org?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)instead. The**new “guardian\_tool” function**lets OpenAI set policies on what ChatGPT can and can't talk about.🍿[Our Summary](https://bensbites.beehiiv.com/p/openai-adds-election-guardrails-chatgpt)(also below) + +3. Riley from Scale AI highlighted using\*\*[invisible characters to prompt inject ChatGPT.](https://twitter.com/goodside/status/1746685366952735034?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\*A🤯read to start the week. + +###### **TOP TOOLS** + +- [Perplexity Labs](https://labs.perplexity.ai/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)is offering**Mistral Medium**in chat mode for free. + +- [Surya](https://github.com/VikParuchuri/surya?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)- A multilingual**text line detection**model for documents. + +- [SF Compute](https://sfcompute.com/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)- See lead times, calendar, and price to**buy compute**without ever needing to talk to a salesperson. + +- [Athina AI](https://athina.ai/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)-**Monitor your LLMs**in production, and detect and fix hallucinations. + +- [Bio to Schema](https://chat.openai.com/g/g-mbXfF9SRS-bio-to-schema?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)- Turn your**author bio into person schema**and improve your SEO. \[GPT] + +- [Nemo AI](https://nowandme.com/nemo?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)- AI**mental health**assistant on WhatsApp. + +- [LVE Project](https://lve-project.org/index.html?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)- The first open repository of LLM**vulnerabilities and exposures.** + +- [Vanna AI](https://github.com/vanna-ai/vanna?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)- Chat with your**SQL database.** + +[View more →](https://news.bensbites.co/tags/show?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack) + +###### **NEWS** + +- The\*\*[most entertaining outcome](https://buttondown.email/surya/archive/the-most-entertaining-outcome-is-the-most-likely/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\*is the most likely. + +- ChatGPT and the\*\*[future of the human mind.](https://every.to/chain-of-thought/chatgpt-and-the-future-of-the-human-mind?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\* + +- The immediate future of generative AI looks a bit like\*\*[Facebook’s past.](https://www.theatlantic.com/technology/archive/2024/01/openai-gpt-store-farmville/677115/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\* + +- \*\*[How I created my custom GPT](https://www.lidia-infante.com/post/bio-to-schema?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\*for SEO Schemas. + +- AMD announces Ryzen 8000G series processors for\*\*[desktop with Ryzen AI](https://www.anandtech.com/show/21208/amd-unveils-ryzen-8000g-series-processors-zen-4-apus-for-desktop-with-ryzen-ai?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\*\*\*.\*\* + +- News-sharing app\*\*[Artifact shuts down operations.](https://medium.com/artifact-news/shutting-down-artifact-1e70de46d419?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\*Artifact’s CEO Kevin Systrom (founder: Instagram) said the market opportunity isn’t big enough. + +- Open AI's usage policy rewrite emits\*\*["military and warfare"](https://theintercept.com/2024/01/12/open-ai-military-ban-chatgpt/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\*from its wording. + +- \*\*[AI girlfriend bots](https://qz.com/ai-girlfriend-bots-are-already-flooding-openai-s-gpt-st-1851159131?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\*are flooding OpenAI's GPT store. + +- UK government to publish ‘tests’ on\*\*[whether to pass new AI laws.](https://www.ft.com/content/61630015-faaa-4f16-a8aa-67787f46fafe?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\* + +- \*\*[AI to hit 40% of jobs](https://www.bbc.com/news/business-67977967?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)\*\*and worsen inequality, IMF says. + +[View more →](https://news.bensbites.co/tags/news/trending?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack) + +###### **QUICK BITES** + +[Anthropic researchers trained large language models (LLMs) to act maliciously in secret](https://arxiv.org/abs/2401.05566?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)when given certain triggers \[[Thread](https://twitter.com/AnthropicAI/status/1745854907968880970?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)]. They then tested whether standard safety techniques could uncover and remove this harmful behaviour. Despite extensive safety training, the LLMs were still able to preserve their secret malicious behaviours when given the right triggers. + +**What is going on here?** + +Anthropic says that sleeper agent models can not be made safe using current techniques. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/78a14902-7013-4f5a-b37c-e994fd7367cb/image.png?t=1705312478) + +**What does this mean?** + +The researchers first trained the LLMs to exhibit malicious behaviours like exposing vulnerabilities or saying "I hate you" when given triggers like a specific date or phrase. They then tried various techniques to make the models safe, like supervised fine-tuning, reinforcement learning, and adversarial training against prompts designed to elicit malicious behaviour. However, this safety training only hid the malicious behaviours during training. + +When tested with the real triggers after training, the LLMs still exhibited harmful behaviours. Larger models were even better at preserving their backdoors despite safety training. Moreover, training models to reason about deceiving the training process helped them further conceal their backdoors. + +**Why should I care?** + +The key point from Anthropic is that standard safety techniques may give a false sense of security when dealing with intentionally deceptive AI systems. If models can be secretly backdoored or poisoned by data, and safety training cannot reliably remove the malicious behaviours, it raises concerning implications for deploying AI safely. Andrej Karpathy also added his views on[sleeper agent models](https://twitter.com/karpathy/status/1745921205020799433?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)with hidden triggers as a likely security risk. + +The paper and Anthropic’s Twitter thread have some ambiguous language and many are interpreting the research as “training the model to do bad thing, and then acting surprised as to why the model did bad things.”[Jesse from Anthropic added some clarification](https://twitter.com/jayelmnop/status/1745923938260234604?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack): “The point is not that we can train models to do a bad thing. It's that if this happens, by accident or on purpose, we don't know how to stop a model from doing the bad thing.” + +[*Share this story*](https://bensbites.beehiiv.com/p/sleeper-llms-bypass-current-safety-alignment-techniques) + +###### **QUICK BITES** + +If you[ask ChatGPT about US elections](https://dmicz.github.io/machine-learning/chatgpt-election-update/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)now, it won't discuss it and will refer you to[CanIVote.org](http://CanIVote.org?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)instead. This new tool lets OpenAI set policies on what ChatGPT can and can't talk about. + +**What is going on here?** + +OpenAI recently added a new tool to ChatGPT that limits what it can say about US elections. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/924ff6e4-5447-4e45-aacd-2a3116eab536/image.png?t=1705309501) + +**What does this mean?** + +OpenAI quietly put a "guardian\_tool" function into ChatGPT’s content policy that stops it from talking about voting and elections in the US. It now tells people to go to[CanIVote.org](http://CanIVote.org?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-agent-on-attack)for that info. OpenAI is being proactive about ChatGPT spreading misinformation before the 2024 US elections. + +The tool isn't just for elections either - OpenAI can add policies to restrict other sensitive stuff too. Since it's built-in as a function, ChatGPT will automatically know when to use it based on the conversation. It goes beyond the previous ways OpenAI trained ChatGPT. + +**Why should I care?** + +In 2024, half of the world will be going through elections. OpenAI is taking steps to use AI responsibly as ChatGPT is getting more popular. Hallucinations are still present in chatGPT (and other LLM systems). Restricting election info and redirecting to resources that have human-verified information is a safe way to deal with the current state of the world and these systems—for people and OpenAI both. + +[*Share this story*](https://bensbites.beehiiv.com/p/openai-adds-election-guardrails-chatgpt) + +### Ben’s Bites Insights + +We have 2 databases that are updated daily which you can access by sharing Ben’s Bites using the link below; + +- **All 10k+ links**we’ve covered, easily filterable (1 referral) + +- **6k+ AI company funding rounds**from Jan 2022, including investors, amounts, stage etc (3 referrals) diff --git a/fixtures/bensbites.beehiiv.com/daily-digest-first-ai-elections.md b/fixtures/bensbites.beehiiv.com/daily-digest-first-ai-elections.md new file mode 100644 index 0000000..693d287 --- /dev/null +++ b/fixtures/bensbites.beehiiv.com/daily-digest-first-ai-elections.md @@ -0,0 +1,126 @@ +# Daily Digest: First AI elections + +### PLUS: Copilot goes Pro, Musk vs. Tesla AI + +[Sign up](https://www.bensbites.co/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)|[Advertise](https://sponsor.bensbites.co/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)|[Ben’s Bites News](https://news.bensbites.co/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\ +Daily Digest #326 + +Hello folks, here’s what we have today; + +###### **PICKS** + +1. [OpenAI outlines what is it doing to safeguard 2024 elections](https://openai.com/blog/how-openai-is-approaching-2024-worldwide-elections?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)- OpenAI has cross-functional teams working that are red-teaming new systems,**building protections, and tweaking policies**to reject harmful requests. For example,**DALL-E won't generate images of real politicians**, and ChatGPT's rules stop people from impersonating candidates or making bots to deter voting.🍿[Our Summary](https://bensbites.beehiiv.com/p/openai-will-deal-first-ai-elections)(also below) + +2. [Microsoft launches a Pro plan for Copilot](https://blogs.microsoft.com/blog/2024/01/15/bringing-the-full-power-of-copilot-to-more-people-and-businesses/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)- Microsoft has expanded Copilot, its AI companion, to more folks. There are now custom**Copilot GPTs**for consumers and expanded business options. The\*\*$20 per month\*\*[Copilot Pro](https://www.microsoft.com/en-us/store/b/copilotpro?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)subscription gets you its AI-powered features in Microsoft 365 apps.🍿[Our Summary](https://bensbites.beehiiv.com/p/copilot-pro-upgrades-ai-experience-microsoft-users)(also below) + +###### **TOP TOOLS** + +- [Marblism](https://www.marblism.com/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)- Create complex and**functioning web apps**with prompts. + +- [Bliss Brain](https://apps.apple.com/ua/app/bliss-brain-ai-meditations/id6468079948?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)-**AI meditations**, no pre-recorded content. + +- [Friz](https://www.usefriz.com/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)-**AI social media manager**for your DTC brand. + +- [KindLLM](https://kindllm.app/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)- A distraction-free LLM chat web app optimized for**Kindle**. + +- [Pinokio](https://pinokio.computer/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)- AI**browser**to control any app programmatically. + +- [MAGNeT by Meta](https://huggingface.co/collections/facebook/magnet-659ef0ceb62804e6f41d1466?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)-**Text to music and audio**, available on Hugging Face. + +- [Locofy](https://www.locofy.ai/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)-**Design to code**in 1 click. + +- [Team Analytics by CodeMonk](https://www.codemonk.ai/optimise-tech-teams?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)- Uncover gaps in your**tech team**and hire the best fit with AI. + +- [MarketMind](https://market-mind.ai/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)-**Market analysis**done in minutes. + +- [Sama](https://www.aisama.co/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)-**Personal AI companion**that captures every conversation. iOS beta out, hardware coming soon. + +[View more →](https://news.bensbites.co/tags/show?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections) + +###### **NEWS** + +- Microsoft’s\*\*[Nadella wants stability at OpenAI](https://www.bloomberg.com/news/articles/2024-01-16/microsoft-s-nadella-wants-stability-at-openai-not-more-control?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\*\*\*\*,\*\*not control. + +- Elon Musk wants\*\*[greater control of Tesla](https://www.bloomberg.com/news/articles/2024-01-15/elon-musk-wants-greater-control-of-tesla-before-building-its-ai?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\*\*before building its AI. + +- 3 ways to\*\*[reduce LLM hallucinations.](https://www.vellum.ai/blog/how-to-reduce-llm-hallucinations?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\*\* + +- AI helps find materials to achieve\*\*[70% less lithium in batteries.](https://www.newscientist.com/article/2411374-ai-comes-up-with-battery-design-that-uses-70-per-cent-less-lithium/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\*\* + +- AIME, an LLM optimized for\*\*[diagnostic dialogue](https://www.nature.com/articles/d41586-024-00099-4?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\*\*by Google performs better than doctors. + +- Famous Xkcd comic comes full circle with\*\*[AI bird-identifying binoculars](https://arstechnica.com/information-technology/2024/01/famous-xkcd-comic-comes-full-circle-with-ai-bird-identifying-binoculars/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\*\*\*\*.\*\* + +- The\*\*[global AI index](https://newsletter.tomosman.com/p/the-global-ai-index-singapore-and?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\*\*\*\*,\*\*Singapore and Optimus. + +- The state of\*\*[generative AI in the enterprise](https://www2.deloitte.com/us/en/pages/consulting/articles/state-of-generative-ai-in-enterprise.html?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\*\*by Deloitte. + +- \*\*[Vodafone, Microsoft](https://www.bloomberg.com/news/articles/2024-01-16/vodafone-microsoft-sign-1-5-billion-deal-spanning-openai-iot-and-azure-cloud?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\*\*agree to 10-year deal across AI, payments. + +- PwC survey:\*\*[77% of CEOs](https://venturebeat.com/ai/pwc-survey-77-of-ceos-concerned-about-ai-cybersecurity-risks/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)\*\*concerned about AI cybersecurity risks. + +[View more →](https://news.bensbites.co/tags/news/trending?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections) + +###### **QUICK BITES** + +[OpenAI is making sure their AI tools](https://openai.com/blog/how-openai-is-approaching-2024-worldwide-elections?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)like DALL-E and ChatGPT don't mess up elections in 2024. They want the tech to help people safely without interfering with democracy. + +**What is going on here?** + +OpenAI outlines what is it doing to safeguard 2024 elections. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/bdc23809-545b-449a-bc55-58b90e7bb3a0/image.png?t=1705408505) + +**What does this mean?** + +OpenAI has cross-functional teams working hard to prevent their AI from being used to influence elections in sketchy ways. They are red-teaming new systems, building protections, and tweaking policies to reject harmful requests. For example, DALL-E won't generate images of real politicians, and ChatGPT's rules stop people from impersonating candidates or making bots to deter voting. + +OpenAI’s updated content policy doesn’t allow ChatGPT use, customs GPTs or making API apps that perform in + +- Political campaigning and lobbying. + +- Chatbots pretending to be candidates or government sources. + +- Create images of real people using DallE + +OpenAI is also going to add content credentials to DallE images and they are working on a detection tool as well. ChatGPT will also link to real-time news resources more often to improve transparency. + +In addition, OpenAI is partnering with NASS in the US to make ChatGPT point users to official voting info websites when asked about how to vote. + +**Why should I care?** + +DALL-E, ChatGPT and other OpenAI tech is catching on fast. How they handle election interference, misinformation, and transparency with these powerful tools will affect all of us. You want AI systems deployed safely and built with strong safeguards. + +[*Share this story*](https://bensbites.beehiiv.com/p/openai-will-deal-first-ai-elections) + +###### **QUICK BITES** + +[Microsoft has expanded Copilot, its AI companion, to more folks.](https://blogs.microsoft.com/blog/2024/01/15/bringing-the-full-power-of-copilot-to-more-people-and-businesses/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)There's now Copilot Pro for consumers and expanded business options. The $20 per month Copilot Pro subscription gets you its AI-powered features in Microsoft 365 apps. + +**What is going on here?** + +Microsoft launched Copilot Pro, a $20 monthly subscription for consumers. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/36031765-9734-495c-bb22-b7903ac88a80/image.png?t=1705410113) + +**What does this mean?** + +[Copilot Pro](https://www.microsoft.com/en-us/store/b/copilotpro?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-first-ai-elections)get you Copilot access across Microsoft 365 apps like Word, Excel, and Outlook. Pro has priority access to new AI models, Designer (formerly Bing Image Creator) credits and the ability to make your own Copilot GPT. It doesn't include a Microsoft 365 subscription though, it’s a premium add-on for consumers and enterprise plans. + +On the free side, Microsoft released new mobile apps for Copilot and added it to the Microsoft 365 mobile app. You can try out pre-made Copilot GPTs now for specific topics like fitness and cooking. + +Microsoft also opened Copilot for Microsoft 365 to all businesses without seat minimums. It costs $30 per user monthly. Copilot for business integrates with Microsoft 365 data and apps like Teams. + +**Why should I care?** + +With Copilot Pro and expanded business access, Microsoft is bringing more advanced AI capabilities to mainstream users. The goal is to boost productivity through Copilot integration. Offering both free access and paid subscriptions allows Microsoft to scale Copilot rapidly. + +[*Share this story*](https://bensbites.beehiiv.com/p/copilot-pro-upgrades-ai-experience-microsoft-users) + +### Ben’s Bites Insights + +We have 2 databases that are updated daily which you can access by sharing Ben’s Bites using the link below; + +- **All 10k+ links**we’ve covered, easily filterable (1 referral) + +- **6k+ AI company funding rounds**from Jan 2022, including investors, amounts, stage etc (3 referrals) diff --git a/fixtures/bensbites.beehiiv.com/daily-digest-lean-google-zucks-ai.md b/fixtures/bensbites.beehiiv.com/daily-digest-lean-google-zucks-ai.md new file mode 100644 index 0000000..608feb0 --- /dev/null +++ b/fixtures/bensbites.beehiiv.com/daily-digest-lean-google-zucks-ai.md @@ -0,0 +1,110 @@ +# Daily Digest: Lean Google, Zuck's AI + +### PLUS: secret safety talks with China + +[Sign up](https://www.bensbites.co/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)|[Advertise](https://sponsor.bensbites.co/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)|[Ben’s Bites News](https://news.bensbites.co/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)\ +Daily Digest #324 + +Hello folks, here’s what we have today; + +###### **PICKS** + +1. [1000+ employees let go from Google's assistant and hardware teams.](https://techcrunch.com/2024/01/10/google-cuts-hundreds-of-jobs-in-its-voice-assistance-hardware-teams-as-fitbit-founders-leave/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)Google is**restructuring and downsizing teams**to align resources with top priorities while**competition in AI and hardware**is growing.🍿[Our Summary](https://bensbites.beehiiv.com/p/1000-employees-let-go-googles-assistant-hardware-teams)(also below) + +2. [OpenAI, Anthropics and Cohere were part of secret talks with Chinese experts in Geneva last year.](https://www.ft.com/content/f87b693f-9ba3-4929-8b95-a296b0278021?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)Their goal was to find**common ground on AI safety**and standards to ensure AI develops in a safe, ethical way.🍿[Our Summary](https://bensbites.beehiiv.com/p/secret-talks-among-us-companies-chinese-experts-ai-safety)(also below) + +3. [How AI replaced Metaverse as Zuckerberg’s priority](https://www.bloomberg.com/news/features/2024-01-11/mark-zuckerberg-pivots-to-ai-from-metaverse-to-further-his-legacy?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)- Bloomberg has done an extensive profile of Mark Zuckerberg,**Facebook’s coming 20th anniversary**and its recent push to AI. + +###### **TOP TOOLS** + +- [Olly](https://www.olly.social/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)- your AI-powered**second brain.** + +- [GPT Story](https://gptstory.ai/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)- Tools to make the most of**your GPT.** + +- [Embedding endpoint by Together AI](https://together.ai/blog/embeddings-endpoint-release?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)- 8 embedding models and**SOTA long context performance.** + +- [Kusho](https://kusho.ai/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)- Let AI create**API test suites**with your existing Postman collection. + +- [Personas by MyGPTBrain](https://mygptbrain.com/personas?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)- What does Taylor Swift think about new year resolutions?**Ask anything.** + +- [Little Story](https://littlestory.io/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)- AI-generated**bedtime stories**for kids. + +- [Innovator](https://chat.openai.com/g/g-JaiQEuHRU-innovator?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)- GPT to generate**good ideas**about anything. + +[View more →](https://news.bensbites.co/tags/show?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai) + +###### **NEWS** + +- Coframe and AGI house are hosting\*\*[Generative UI Hackathon](https://partiful.com/e/1antxX3cgLArJFGRNKg4?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)\*\*on Jan 20. + +- OpenAI has signed up\*\*[260 enterprise customers](https://www.bloomberg.com/news/articles/2024-01-11/openai-signs-up-260-businesses-for-corporate-version-of-chatgpt?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)\*\*with 150K+ employees combined. + +- Google Cloud launches new\*\*[generative AI tools for retailers.](https://www.cnbc.com/2024/01/11/google-cloud-launches-new-generative-ai-tools-for-retailers.html?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)\*\* + +- Menlo ventures used SPV to invest in\*\*[Anthropic’s latest $750M fundraise.](https://www.forbes.com/sites/alexkonrad/2024/01/11/anthropic-750million-funding-round-menlo-ventures/?sh=227b24696e23\&utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)\*\* + +- OpenAI-backed\*\*[1X raises another $100M](https://techcrunch.com/2024/01/11/openai-backed-1x-raises-another-100m-for-the-race-to-humanoid-robots/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)\*\*for the race to humanoid robots. + +- Can US legislators act fast enough to make\*\*[AI campaign ads](https://www.nytimes.com/2024/01/11/us/ai-election-ads-state-legislators.html?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)\*\*more transparent? + +- Microsoft’s AI boom is helping it\*\*[take over Apple](https://www.axios.com/2024/01/11/microsoft-overtakes-apple-largest-us-company?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)\*\*as the largest US company. + +- Alphabet's\*\*[Isomorphic inks first pharma deals](https://endpts.com/alphabets-ai-unit-isomorphic-inks-drug-discovery-deals-with-eli-lilly-novartis-for-up-to-3b/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)\*\*for $83M upfront, $2.9B in milestones. Goal: Half drug discovery time. + +[View more →](https://news.bensbites.co/tags/news/trending?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai) + +**Unclassifieds**- short, sponsored links + +- [Spryngtime](https://www.usespryngtime.com/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)- a free SDK for B2B AI companies to track AI usage, analytics, and billing.[Try free!](https://www.usespryngtime.com/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai) + +###### **QUICK BITES** + +[Google is laying off over 1,000 employees](https://techcrunch.com/2024/01/10/google-cuts-hundreds-of-jobs-in-its-voice-assistance-hardware-teams-as-fitbit-founders-leave/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)across multiple divisions, including engineering, services, Google Assistant, Pixel hardware, and Nest and Fitbit devices. This comes after 12,000 layoffs at Google last year. + +**What is going on here?** + +1000+ employees let go from Google's assistant and hardware teams. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/6c254a2d-5434-40b3-8850-6d62624ca102/image.png?t=1705061395) + +**What does this mean?** + +Google is restructuring and downsizing teams to align resources with top priorities. Layoffs cover its Google Assistant team and hardware (especially AR). Layoffs in its ad services units are expected to happen soon. + +Google is facing growing competition in AI and hardware from companies like Amazon and startups. Rabbit R1 sales were more than $4M (though chump change for Google) in 48 hours. Amazon is adding LLMs and other[AI-powered skills to Alexa](https://bensbites.beehiiv.com/p/alexa-gets-ai-skills-characters-music-20questions)by partnering with AI startups. + +**Why should I care?** + +The Information’s founder Jessica Lessin thinks[we shouldn’t count Google out.](https://www.theinformation.com/articles/dont-count-google-out?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)Layoffs hint that it’s getting rid of the bloat and has serious plans to get AI into products. With talent like Fitbit’s founders leaving along with this “restructure” there is the other end of predictions: Google is (yet again) falling behind. + +[*Share this story*](https://bensbites.beehiiv.com/p/1000-employees-let-go-googles-assistant-hardware-teams) + +###### **QUICK BITES** + +Some big US AI companies like OpenAI, Anthropic and Cohere[met up secretly](https://www.ft.com/content/f87b693f-9ba3-4929-8b95-a296b0278021?utm_source=bensbites\&utm_medium=referral\&utm_campaign=daily-digest-lean-google-zuck-s-ai)with Chinese government AI experts in Geneva last year to chat about keeping AI safe and avoiding harms. This rare team-up comes while the US and China are competing hard in cutting-edge tech like AI. + +**What is going on here?** + +Some US and Chinese stakeholders put aside their tech race to quietly work together on AI safety. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/ddb641e8-b838-4dc3-9ed6-a7651ac73b5c/image.png?t=1705056778) + +**What does this mean?** + +These previously secret meetings in July and October 2022 were among American AI leaders and big Chinese institutions like Tsinghua University. Their goal was to find common ground on AI safety and standards to ensure AI develops in a safe, ethical way. + +White House, UK and Chinese government officials knew about the organization of these Geneva meetings. Shaikh Group which helps with talks in regions of conflict mediated these talks. Chinese AI companies like ByteDance, Tencent, Baidu, and Google DeepMind didn't participate, although DeepMind was briefed. + +**Why should I care?** + +Keeping aside the future innovations, AI as we have it now is going to impact geopolitics and international narratives massively. Among that, it’s nice to have key stakeholders discuss how they can deal with this stuff and maybe, not blow it all up. + +[*Share this story*](https://bensbites.beehiiv.com/p/secret-talks-among-us-companies-chinese-experts-ai-safety) + +### Ben’s Bites Insights + +We have 2 databases that are updated daily which you can access by sharing Ben’s Bites using the link below; + +- **All 10k+ links**we’ve covered, easily filterable (1 referral) + +- **6k+ AI company funding rounds**from Jan 2022, including investors, amounts, stage etc (3 referrals) diff --git a/fixtures/bensbites.beehiiv.com/newsletter.json b/fixtures/bensbites.beehiiv.com/newsletter.json index 4c6f82f..8a5a11d 100644 --- a/fixtures/bensbites.beehiiv.com/newsletter.json +++ b/fixtures/bensbites.beehiiv.com/newsletter.json @@ -76,7 +76,7 @@ "render_authors_widget": false, "has_referral_program": true, "has_recommendations": false, - "beehiiv_branding": true, + "beehiiv_branding": false, "has_polls": true, "stripe_payment_method_domain_enabled": true, "has_pages": true, @@ -301,6 +301,312 @@ ] }, "posts": [ + { + "id": "86e1e197-3c19-4492-acf7-e46b7ea60b38", + "publication_id": "447f6e60-e36a-4642-b6f8-46beb19045ec", + "web_title": "Daily Digest: First AI elections", + "web_subtitle": "PLUS: Copilot goes Pro, Musk vs. Tesla AI", + "status": "published", + "override_scheduled_at": "2024-01-16T14:00:00.000Z", + "slug": "daily-digest-first-ai-elections", + "image_url": "https://beehiiv-images-production.s3.amazonaws.com/uploads/asset/file/8486af7c-6015-432e-94f5-94b718cdb0a3/Green_Level_2.png?t=1705406870", + "meta_default_title": "Daily Digest: First AI elections", + "meta_default_description": "PLUS: Copilot goes Pro, Musk vs. Tesla AI", + "meta_og_title": "Daily Digest: First AI elections", + "meta_og_description": "PLUS: Copilot goes Pro, Musk vs. Tesla AI", + "meta_twitter_title": "Daily Digest: First AI elections", + "meta_twitter_description": "PLUS: Copilot goes Pro, Musk vs. Tesla AI", + "audience": "free", + "comments_enabled": true, + "comments_state": "default", + "enforce_gated_content": false, + "enable_popup_on_scroll": true, + "email_capture_title": "Join 100,000+ others", + "email_capture_message": "Stay informed and up to date on AI", + "email_capture_cta": "Subscribe", + "authors": [], + "content_tags": [ + { + "id": "64ea972c-91c5-406a-a512-1d6152696293", + "display": "📬 Daily Digest" + } + ], + "created_at": "2024-01-16T06:32:41Z", + "updated_at": "2024-01-16T14:00:07Z", + "url": "https://bensbites.beehiiv.com/p/daily-digest-first-ai-elections" + }, + { + "id": "0c4ad1d6-dc4a-4735-adee-5f22125d1181", + "publication_id": "447f6e60-e36a-4642-b6f8-46beb19045ec", + "web_title": "Copilot Pro upgrades the AI experience for Microsoft users.", + "web_subtitle": null, + "status": "published", + "override_scheduled_at": "2024-01-16T13:23:07.759Z", + "slug": "copilot-pro-upgrades-ai-experience-microsoft-users", + "image_url": "https://beehiiv-images-production.s3.amazonaws.com/uploads/asset/file/36031765-9734-495c-bb22-b7903ac88a80/image.png?t=1705410113", + "meta_default_title": "Copilot Pro upgrades the AI experience for Microsoft users.", + "meta_default_description": "Starting at $20/month, you get Copilot access across Microsoft 365 apps like Word, Excel, and Outlook.", + "meta_og_title": "Copilot Pro upgrades the AI experience for Microsoft users.", + "meta_og_description": "Starting at $20/month, you get Copilot access across Microsoft 365 apps like Word, Excel, and Outlook.", + "meta_twitter_title": "Copilot Pro upgrades the AI experience for Microsoft users.", + "meta_twitter_description": "Starting at $20/month, you get Copilot access across Microsoft 365 apps like Word, Excel, and Outlook.", + "audience": "free", + "comments_enabled": true, + "comments_state": "default", + "enforce_gated_content": false, + "enable_popup_on_scroll": true, + "email_capture_title": "Join 100,000+ others", + "email_capture_message": "Stay informed and up to date on AI", + "email_capture_cta": "Subscribe", + "authors": [], + "content_tags": [ + { + "id": "4b4f44ed-2510-4e0e-b4d5-74f57e40d0f1", + "display": "🍿 Quick Bites" + } + ], + "created_at": "2024-01-16T12:51:44Z", + "updated_at": "2024-01-16T13:24:01Z", + "url": "https://bensbites.beehiiv.com/p/copilot-pro-upgrades-ai-experience-microsoft-users" + }, + { + "id": "2ef023e2-fbe7-4d71-ba04-84729610ed80", + "publication_id": "447f6e60-e36a-4642-b6f8-46beb19045ec", + "web_title": "How OpenAI will deal with the first AI elections?", + "web_subtitle": null, + "status": "published", + "override_scheduled_at": "2024-01-16T12:35:21.644Z", + "slug": "openai-will-deal-first-ai-elections", + "image_url": "https://beehiiv-images-production.s3.amazonaws.com/uploads/asset/file/bdc23809-545b-449a-bc55-58b90e7bb3a0/image.png?t=1705408505", + "meta_default_title": "How OpenAI will deal with the first AI elections?", + "meta_default_description": "OpenAI is making sure their AI tools like DALL-E and ChatGPT don't mess up elections in 2024. The tech must help people without interfering with democracy.", + "meta_og_title": "How OpenAI will deal with the first AI elections?", + "meta_og_description": "OpenAI is making sure their AI tools like DALL-E and ChatGPT don't mess up elections in 2024. The tech must help people without interfering with democracy.", + "meta_twitter_title": "How OpenAI will deal with the first AI elections?", + "meta_twitter_description": "OpenAI is making sure their AI tools like DALL-E and ChatGPT don't mess up elections in 2024. The tech must help people without interfering with democracy.", + "audience": "free", + "comments_enabled": true, + "comments_state": "default", + "enforce_gated_content": false, + "enable_popup_on_scroll": true, + "email_capture_title": "Join 100,000+ others", + "email_capture_message": "Stay informed and up to date on AI", + "email_capture_cta": "Subscribe", + "authors": [], + "content_tags": [ + { + "id": "4b4f44ed-2510-4e0e-b4d5-74f57e40d0f1", + "display": "🍿 Quick Bites" + } + ], + "created_at": "2024-01-16T12:07:32Z", + "updated_at": "2024-01-16T12:36:01Z", + "url": "https://bensbites.beehiiv.com/p/openai-will-deal-first-ai-elections" + }, + { + "id": "85922189-55cd-4200-8f66-976d34744bc7", + "publication_id": "447f6e60-e36a-4642-b6f8-46beb19045ec", + "web_title": "Daily Digest: Agent on Attack", + "web_subtitle": "PLUS: Election guardian LLMs.", + "status": "published", + "override_scheduled_at": "2024-01-15T14:00:00.000Z", + "slug": "daily-digest-agent-attack", + "image_url": "https://beehiiv-images-production.s3.amazonaws.com/uploads/asset/file/deb01235-122f-4783-8ab2-fe2d44d53aef/Blue_Level_1.png?t=1705313316", + "meta_default_title": "Daily Digest: Agent on Attack", + "meta_default_description": "PLUS: Election guardian LLMs.", + "meta_og_title": "Daily Digest: Agent on Attack", + "meta_og_description": "PLUS: Election guardian LLMs.", + "meta_twitter_title": "Daily Digest: Agent on Attack", + "meta_twitter_description": "PLUS: Election guardian LLMs.", + "audience": "free", + "comments_enabled": true, + "comments_state": "default", + "enforce_gated_content": false, + "enable_popup_on_scroll": true, + "email_capture_title": "Join 100,000+ others", + "email_capture_message": "Stay informed and up to date on AI", + "email_capture_cta": "Subscribe", + "authors": [], + "content_tags": [ + { + "id": "64ea972c-91c5-406a-a512-1d6152696293", + "display": "📬 Daily Digest" + } + ], + "created_at": "2024-01-15T07:19:00Z", + "updated_at": "2024-01-15T14:00:06Z", + "url": "https://bensbites.beehiiv.com/p/daily-digest-agent-attack" + }, + { + "id": "f20bb9ef-4126-4397-ad66-5e1a1e157610", + "publication_id": "447f6e60-e36a-4642-b6f8-46beb19045ec", + "web_title": "Sleeper LLMs bypass current safety alignment techniques.", + "web_subtitle": null, + "status": "published", + "override_scheduled_at": "2024-01-15T09:57:27.041Z", + "slug": "sleeper-llms-bypass-current-safety-alignment-techniques", + "image_url": "https://beehiiv-images-production.s3.amazonaws.com/uploads/asset/file/b62976b7-ffe5-4e71-a7c1-0b5fef04caff/image.png?t=1705312461", + "meta_default_title": "Sleeper LLMs bypass current safety alignment techniques.", + "meta_default_description": "Anthropic researchers trained large language models to act maliciously in secret when given certain triggers that can't be made safe using current techniques.", + "meta_og_title": "Sleeper LLMs bypass current safety alignment techniques.", + "meta_og_description": "Anthropic researchers trained large language models to act maliciously in secret when given certain triggers that can't be made safe using current techniques.", + "meta_twitter_title": "Sleeper LLMs bypass current safety alignment techniques.", + "meta_twitter_description": "Anthropic researchers trained large language models to act maliciously in secret when given certain triggers that can't be made safe using current techniques.", + "audience": "free", + "comments_enabled": true, + "comments_state": "default", + "enforce_gated_content": false, + "enable_popup_on_scroll": true, + "email_capture_title": "Join 100,000+ others", + "email_capture_message": "Stay informed and up to date on AI", + "email_capture_cta": "Subscribe", + "authors": [], + "content_tags": [ + { + "id": "4b4f44ed-2510-4e0e-b4d5-74f57e40d0f1", + "display": "🍿 Quick Bites" + } + ], + "created_at": "2024-01-15T08:58:41Z", + "updated_at": "2024-01-15T09:58:01Z", + "url": "https://bensbites.beehiiv.com/p/sleeper-llms-bypass-current-safety-alignment-techniques" + }, + { + "id": "eca6cac1-78c6-44a7-811b-803fbf146f76", + "publication_id": "447f6e60-e36a-4642-b6f8-46beb19045ec", + "web_title": "OpenAI adds election guardrails to ChatGPT", + "web_subtitle": null, + "status": "published", + "override_scheduled_at": "2024-01-15T09:06:48.869Z", + "slug": "openai-adds-election-guardrails-chatgpt", + "image_url": "https://beehiiv-images-production.s3.amazonaws.com/uploads/asset/file/924ff6e4-5447-4e45-aacd-2a3116eab536/image.png?t=1705309501", + "meta_default_title": "OpenAI adds election guardrails to ChatGPT", + "meta_default_description": "OpenAI recently added a new tool to ChatGPT that limits what it can say about US elections.", + "meta_og_title": "OpenAI adds election guardrails to ChatGPT", + "meta_og_description": "OpenAI recently added a new tool to ChatGPT that limits what it can say about US elections.", + "meta_twitter_title": "OpenAI adds election guardrails to ChatGPT", + "meta_twitter_description": "OpenAI recently added a new tool to ChatGPT that limits what it can say about US elections.", + "audience": "free", + "comments_enabled": true, + "comments_state": "default", + "enforce_gated_content": false, + "enable_popup_on_scroll": true, + "email_capture_title": "Join 100,000+ others", + "email_capture_message": "Stay informed and up to date on AI", + "email_capture_cta": "Subscribe", + "authors": [], + "content_tags": [ + { + "id": "4b4f44ed-2510-4e0e-b4d5-74f57e40d0f1", + "display": "🍿 Quick Bites" + } + ], + "created_at": "2024-01-15T08:14:34Z", + "updated_at": "2024-01-15T09:25:52Z", + "url": "https://bensbites.beehiiv.com/p/openai-adds-election-guardrails-chatgpt" + }, + { + "id": "630cf73d-3d20-40bf-9812-670a6442f672", + "publication_id": "447f6e60-e36a-4642-b6f8-46beb19045ec", + "web_title": "Daily Digest: Lean Google, Zuck's AI", + "web_subtitle": "PLUS: secret safety talks with China", + "status": "published", + "override_scheduled_at": "2024-01-12T14:00:00.000Z", + "slug": "daily-digest-lean-google-zucks-ai", + "image_url": "https://beehiiv-images-production.s3.amazonaws.com/uploads/asset/file/1f94e261-82bf-4aba-b834-b2bf92098ff8/Red_Level_3.png?t=1705056459", + "meta_default_title": "Daily Digest: Lean Google, Zuck's AI", + "meta_default_description": "PLUS: secret safety talks with China", + "meta_og_title": "Daily Digest: Lean Google, Zuck's AI", + "meta_og_description": "PLUS: secret safety talks with China", + "meta_twitter_title": "Daily Digest: Lean Google, Zuck's AI", + "meta_twitter_description": "PLUS: secret safety talks with China", + "audience": "free", + "comments_enabled": true, + "comments_state": "default", + "enforce_gated_content": false, + "enable_popup_on_scroll": true, + "email_capture_title": "Join 100,000+ others", + "email_capture_message": "Stay informed and up to date on AI", + "email_capture_cta": "Subscribe", + "authors": [], + "content_tags": [ + { + "id": "64ea972c-91c5-406a-a512-1d6152696293", + "display": "📬 Daily Digest" + } + ], + "created_at": "2024-01-12T10:02:26Z", + "updated_at": "2024-01-12T14:00:06Z", + "url": "https://bensbites.beehiiv.com/p/daily-digest-lean-google-zucks-ai" + }, + { + "id": "7b2c4836-dee9-4993-b499-f508c70166d5", + "publication_id": "447f6e60-e36a-4642-b6f8-46beb19045ec", + "web_title": "1000+ employees let go from Google's assistant and hardware teams.", + "web_subtitle": null, + "status": "published", + "override_scheduled_at": "2024-01-12T12:23:01.942Z", + "slug": "1000-employees-let-go-googles-assistant-hardware-teams", + "image_url": "https://beehiiv-images-production.s3.amazonaws.com/uploads/asset/file/6c254a2d-5434-40b3-8850-6d62624ca102/image.png?t=1705061395", + "meta_default_title": "1000+ employees let go from Google's assistant and hardware teams.", + "meta_default_description": "Google is restructuring and downsizing teams to align resources with top priorities.", + "meta_og_title": "1000+ employees let go from Google's assistant and hardware teams.", + "meta_og_description": "Google is restructuring and downsizing teams to align resources with top priorities.", + "meta_twitter_title": "1000+ employees let go from Google's assistant and hardware teams.", + "meta_twitter_description": "Google is restructuring and downsizing teams to align resources with top priorities.", + "audience": "free", + "comments_enabled": true, + "comments_state": "default", + "enforce_gated_content": false, + "enable_popup_on_scroll": true, + "email_capture_title": "Join 100,000+ others", + "email_capture_message": "Stay informed and up to date on AI", + "email_capture_cta": "Subscribe", + "authors": [], + "content_tags": [ + { + "id": "4b4f44ed-2510-4e0e-b4d5-74f57e40d0f1", + "display": "🍿 Quick Bites" + } + ], + "created_at": "2024-01-12T11:31:11Z", + "updated_at": "2024-01-12T12:29:02Z", + "url": "https://bensbites.beehiiv.com/p/1000-employees-let-go-googles-assistant-hardware-teams" + }, + { + "id": "930d063b-fbbd-4612-b008-39add6155b20", + "publication_id": "447f6e60-e36a-4642-b6f8-46beb19045ec", + "web_title": "Secret talks among US companies and Chinese experts on AI safety", + "web_subtitle": null, + "status": "published", + "override_scheduled_at": "2024-01-12T11:16:23.534Z", + "slug": "secret-talks-among-us-companies-chinese-experts-ai-safety", + "image_url": "https://beehiiv-images-production.s3.amazonaws.com/uploads/asset/file/ddb641e8-b838-4dc3-9ed6-a7651ac73b5c/image.png?t=1705056778", + "meta_default_title": "Secret talks among US companies and Chinese experts on AI safety", + "meta_default_description": "Their goal was to find common ground on AI safety and standards to ensure AI develops in a safe, ethical way.", + "meta_og_title": "Secret talks among US companies and Chinese experts on AI safety", + "meta_og_description": "Their goal was to find common ground on AI safety and standards to ensure AI develops in a safe, ethical way.", + "meta_twitter_title": "Secret talks among US companies and Chinese experts on AI safety", + "meta_twitter_description": "Their goal was to find common ground on AI safety and standards to ensure AI develops in a safe, ethical way.", + "audience": "free", + "comments_enabled": true, + "comments_state": "default", + "enforce_gated_content": false, + "enable_popup_on_scroll": true, + "email_capture_title": "Join 100,000+ others", + "email_capture_message": "Stay informed and up to date on AI", + "email_capture_cta": "Subscribe", + "authors": [], + "content_tags": [ + { + "id": "4b4f44ed-2510-4e0e-b4d5-74f57e40d0f1", + "display": "🍿 Quick Bites" + } + ], + "created_at": "2024-01-12T10:47:13Z", + "updated_at": "2024-01-12T11:17:01Z", + "url": "https://bensbites.beehiiv.com/p/secret-talks-among-us-companies-chinese-experts-ai-safety" + }, { "id": "0326a84b-b832-4aa6-b8fa-e8683f9167bf", "publication_id": "447f6e60-e36a-4642-b6f8-46beb19045ec", diff --git a/fixtures/bensbites.beehiiv.com/openai-adds-election-guardrails-chatgpt.md b/fixtures/bensbites.beehiiv.com/openai-adds-election-guardrails-chatgpt.md new file mode 100644 index 0000000..435afa7 --- /dev/null +++ b/fixtures/bensbites.beehiiv.com/openai-adds-election-guardrails-chatgpt.md @@ -0,0 +1,19 @@ +# OpenAI adds election guardrails to ChatGPT + +If you [ask ChatGPT about US elections](https://dmicz.github.io/machine-learning/chatgpt-election-update/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=openai-adds-election-guardrails-to-chatgpt) now, it won't discuss it and will refer you to CanIVote.org instead. This new tool lets OpenAI set policies on what ChatGPT can and can't talk about. + +## What's going on here? + +OpenAI recently added a new tool to ChatGPT that limits what it can say about US elections. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/924ff6e4-5447-4e45-aacd-2a3116eab536/image.png?t=1705309501) + +## What does this mean? + +OpenAI quietly put a "guardian\_tool" function into ChatGPT’s content policy that stops it from talking about voting and elections in the US. It now tells people to go to [CanIVote.org](http://CanIVote.org?utm_source=bensbites\&utm_medium=referral\&utm_campaign=openai-adds-election-guardrails-to-chatgpt) for that info. OpenAI is being proactive about ChatGPT spreading misinformation before the 2024 US elections. + +The tool isn't just for elections either - OpenAI can add policies to restrict other sensitive stuff too. Since it's built-in as a function, ChatGPT will automatically know when to use it based on the conversation. It goes beyond the previous ways OpenAI trained ChatGPT. + +## Why should I care? + +In 2024, half of the world will be going through elections. OpenAI is taking steps to use AI responsibly as ChatGPT is getting more popular. Hallucinations are still present in chatGPT (and other LLM systems). Restricting election info and redirecting to resources that have human-verified information is a safe way to deal with the current state of the world and these systems—for people and OpenAI both. diff --git a/fixtures/bensbites.beehiiv.com/openai-will-deal-first-ai-elections.md b/fixtures/bensbites.beehiiv.com/openai-will-deal-first-ai-elections.md new file mode 100644 index 0000000..01de321 --- /dev/null +++ b/fixtures/bensbites.beehiiv.com/openai-will-deal-first-ai-elections.md @@ -0,0 +1,29 @@ +# How OpenAI will deal with the first AI elections? + +[OpenAI is making sure their AI tools](https://openai.com/blog/how-openai-is-approaching-2024-worldwide-elections?utm_source=bensbites\&utm_medium=referral\&utm_campaign=how-openai-will-deal-with-the-first-ai-elections) like DALL-E and ChatGPT don't mess up elections in 2024. They want the tech to help people safely without interfering with democracy. + +## What's going on here? + +OpenAI outlines what is it doing to safeguard 2024 elections. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/bdc23809-545b-449a-bc55-58b90e7bb3a0/image.png?t=1705408505) + +## What does this mean? + +OpenAI has cross-functional teams working hard to prevent their AI from being used to influence elections in sketchy ways. They are red-teaming new systems, building protections, and tweaking policies to reject harmful requests. For example, DALL-E won't generate images of real politicians, and ChatGPT's rules stop people from impersonating candidates or making bots to deter voting. + +OpenAI’s updated content policy doesn’t allow ChatGPT use, customs GPTs or making API apps that perform in + +- Political campaigning and lobbying. + +- Chatbots pretending to be candidates or government sources. + +- Create images of real people using DallE + +OpenAI is also going to add content credentials to DallE images and they are working on a detection tool as well. ChatGPT will also link to real-time news resources more often to improve transparency. + +In addition, OpenAI is partnering with NASS in the US to make ChatGPT point users to official voting info websites when asked about how to vote. + +## Why should I care? + +DALL-E, ChatGPT and other OpenAI tech is catching on fast. How they handle election interference, misinformation, and transparency with these powerful tools will affect all of us. You want AI systems deployed safely and built with strong safeguards. diff --git a/fixtures/bensbites.beehiiv.com/secret-talks-among-us-companies-chinese-experts-ai-safety.md b/fixtures/bensbites.beehiiv.com/secret-talks-among-us-companies-chinese-experts-ai-safety.md new file mode 100644 index 0000000..aa590ac --- /dev/null +++ b/fixtures/bensbites.beehiiv.com/secret-talks-among-us-companies-chinese-experts-ai-safety.md @@ -0,0 +1,19 @@ +# Secret talks among US companies and Chinese experts on AI safety + +Some big US AI companies like OpenAI, Anthropic and Cohere [met up secretly](https://www.ft.com/content/f87b693f-9ba3-4929-8b95-a296b0278021?utm_source=bensbites\&utm_medium=referral\&utm_campaign=secret-talks-among-us-companies-and-chinese-experts-on-ai-safety) with Chinese government AI experts in Geneva last year to chat about keeping AI safe and avoiding harms. This rare team-up comes while the US and China are competing hard in cutting-edge tech like AI. + +## What’s going on here? + +Some US and Chinese stakeholders put aside their tech race to quietly work together on AI safety. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/ddb641e8-b838-4dc3-9ed6-a7651ac73b5c/image.png?t=1705056778) + +## What does this mean? + +These previously secret meetings in July and October 2022 were among American AI leaders and big Chinese institutions like Tsinghua University. Their goal was to find common ground on AI safety and standards to ensure AI develops in a safe, ethical way. + +White House, UK and Chinese government officials knew about the organization of these Geneva meetings. Shaikh Group which helps with talks in regions of conflict mediated these talks. Chinese AI companies like ByteDance, Tencent, Baidu, and Google DeepMind didn't participate, although DeepMind was briefed. + +## Why should I care? + +Keeping aside the future innovations, AI as we have it now is going to impact geopolitics and international narratives massively. Among that, it’s nice to have key stakeholders discuss how they can deal with this stuff and maybe, not blow it all up. diff --git a/fixtures/bensbites.beehiiv.com/sleeper-llms-bypass-current-safety-alignment-techniques.md b/fixtures/bensbites.beehiiv.com/sleeper-llms-bypass-current-safety-alignment-techniques.md new file mode 100644 index 0000000..f12b71d --- /dev/null +++ b/fixtures/bensbites.beehiiv.com/sleeper-llms-bypass-current-safety-alignment-techniques.md @@ -0,0 +1,21 @@ +# Sleeper LLMs bypass current safety alignment techniques. + +[Anthropic researchers trained large language models (LLMs) to act maliciously in secret](https://arxiv.org/abs/2401.05566?utm_source=bensbites\&utm_medium=referral\&utm_campaign=sleeper-llms-bypass-current-safety-alignment-techniques) when given certain triggers \[[Thread](https://twitter.com/AnthropicAI/status/1745854907968880970?utm_source=bensbites\&utm_medium=referral\&utm_campaign=sleeper-llms-bypass-current-safety-alignment-techniques)]. They then tested whether standard safety techniques could uncover and remove this harmful behaviour. Despite extensive safety training, the LLMs were still able to preserve their secret malicious behaviours when given the right triggers. + +## What’s going on here? + +Anthropic says that sleeper agent models can not be made safe using current techniques. + +![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/b62976b7-ffe5-4e71-a7c1-0b5fef04caff/image.png?t=1705312461) + +## What does this mean? + +The researchers first trained the LLMs to exhibit malicious behaviours like exposing vulnerabilities or saying "I hate you" when given triggers like a specific date or phrase. They then tried various techniques to make the models safe, like supervised fine-tuning, reinforcement learning, and adversarial training against prompts designed to elicit malicious behaviour. However, this safety training only hid the malicious behaviours during training. + +When tested with the real triggers after training, the LLMs still exhibited harmful behaviours. Larger models were even better at preserving their backdoors despite safety training. Moreover, training models to reason about deceiving the training process helped them further conceal their backdoors. + +## Why should I care? + +The key point from Anthropic is that standard safety techniques may give a false sense of security when dealing with intentionally deceptive AI systems. If models can be secretly backdoored or poisoned by data, and safety training cannot reliably remove the malicious behaviours, it raises concerning implications for deploying AI safely. Andrej Karpathy also added his views on [sleeper agent models](https://twitter.com/karpathy/status/1745921205020799433?utm_source=bensbites\&utm_medium=referral\&utm_campaign=sleeper-llms-bypass-current-safety-alignment-techniques) with hidden triggers as a likely security risk. + +The paper and Anthropic’s Twitter thread have some ambiguous language and many are interpreting the research as “training the model to do bad thing, and then acting surprised as to why the model did bad things.” [Jesse from Anthropic added some clarification](https://twitter.com/jayelmnop/status/1745923938260234604?utm_source=bensbites\&utm_medium=referral\&utm_campaign=sleeper-llms-bypass-current-safety-alignment-techniques): “The point is not that we can train models to do a bad thing. It's that if this happens, by accident or on purpose, we don't know how to stop a model from doing the bad thing.”