Advertise With Us | Join Discord Community

Hey friends 👋 ,

Happy Sunday my fellow AI champions. It’s an absolute pleasure to have you here once again. Whilst it’s been a massive week for AI, we’re here to cut Through The Noise (a little on the nose, I know), and deliver only the most important AI signal from > 500 pieces of AI-related content. I’ve just brewed my last cup of Earl Grey for the afternoon, so let’s get to it!

…………………………………………………………………………….

🤖 AI Summary

This week’s newsletter, summarised with humour, by ChatGPT in 3 sentences:

Anthropic announced a 4-year, $5 billion plan to build a "frontier model" called "Claude-Next," which is ten times more capable than today's most powerful AI. Joe Biden met with the council of advisors to discuss risks and opportunities in AI advancements and unveiled guidelines for averting harm. MetaAI releases SegmentAnything, an AI model that can cut out any object, in any image, while LumaAI allows you to capture any 3D object and reproduce it with unmatched photorealistic qualities, perhaps useful for super-imposing furniture in your new home.

…………………………………………………………………………….

Today's Through the Noise is brought to you by The Diff.

Looking for a newsletter that's more than just surface-level news?

Check out The Diff, it's like a deep dive into a pool of insightful analysis and commentary ranging from in-depth company profiles to strategy and macroeconomics.

It'll keep you ahead of the pack, on top of your game, and leave you with top-tier conversation for your next dinner party.

Join the ranks of 50,000 founders, VCs, hedge fund managers and others reading the Diff today.

…………………………………………………………………………….

⛏ 5 News Nuggets

• 🔗 Link: Anthropic announce a 4-year, 5 billion-dollar plan to overtake OpenAI. The company plans to build a "frontier model" called "Claude-Next," which is ten times more capable than today's most powerful AI and requires $1 billion in spending over the next 18 months. Looks like the position for an industry leader is heating up.

• 🔗 Link: Joe Biden meets with the council of advisors to discuss risks and opportunities in AI advancements. Citing potential risks to society, the economy and national security—Biden unveiled a set of guidelines aimed at averting harm by protecting children and curbing mass data collection. When asked if AI is dangerous, Biden replies: “It remains to be seen. Could be.”

• 🔗 Link: Bloomberg launches BloombergGPT, an AI model designed for financial NLP (Natural Language Processing) tasks. With 50 billion parameters, BloombergGPT tackles sentiment analysis, named entity recognition, news classification, and question answering. Who knows, maybe it will be the key to predicting the next big stock market crash or at least help us sound smarter about finance.

• 🔗 Link: Y-combinator’s Demo Day is underway, with 60 out of the 273 software companies being AI startups. More than half of the 60 are generative—which builds “ChatGPT for X”. The majority target streamlined workflows as their USP, which begs the question, will we see a resurgence in productivity gurus? (AI-generated perhaps).

• 🔗 Link: India embraces AI, and shuns regulation. The nation has identified the sector as “significant and strategic” and has no plans to regulate growth in the South Asian market. The stance comes at a time of increased regulatory scrutiny, solidifying their commitment to AI’s growth & development.

…………………………………………………………………………….

😎 4 Cool Creations

• 🔗 Link: AutoGPT showcases the capabilities of GPT-4—increasing a firm’s net worth through reasoning and autonomous execution. The open-source project is run by donors & volunteers; further illustrating the collaborative nature of many AI projects.

• 🔗 Link: Engage AI empowers LinkedIn power users to write insightful comments, breaking the ice and building relationships with prospects in a fraction of the time. I’ve been using this recently and found tremendous value.

• 🔗 Link: BanterAI lets you talk to your favourite celebrity over your web browser. Using real-time speech recognition technology and AI-generated avatars—such as Steve Jobs or Elon Musk—you can have a conversation with your biggest idols. I had a lot of fun asking Steve Jobs menial questions—such as, what should I have for dinner?

• 🔗 Link: Toolkit lets you generate your own plug-ins using AI. From just a simple description, you can generate plug-ins to download YouTube videos, retrieve articles, and analyse stock market data.

…………………………………………………………………………….

🔦 3 Startup Spotlights

• 🔗 Link: Quantexa raises $129M in funding to reach a $1.8B valuation. The UK-based AI firm has core products in risk & compliance—detecting money laundering, verifying user identities & carrying out financial investigations. Without reaching profitability, they’ve managed to consistently raise funds over the past 5 years.

• 🔗 Link: Fourthline secures $54M in all-equity funding through Finch Capital & other undisclosed investors. The Amsterdam-based start-up focuses on identity verification, compliance with anti-money laundering rules, and fraud prevention. This investment takes their total funding to $70m after a yearly growth rate of 80% over the past 5 years.

• 🔗 Link: Amazon announces $300K in prize funding for 10 generative AI startups through an accelerator programme. The programme provides access to AI models & tools, go-to-market advice, and more (applications are open for the next 2 weeks). What would you build?

…………………………………………………………………………….

🐥 2 Top Tweets

• 🔗 Link: Meta releases SegmentAnything—a new AI model that can cut out any object, in any image. The model is promptable, meaning you can use text-to-speech to automatically recognise specific objects. In demo mode, the program identifies cats within an image, rapidly picks out objects within a home, and blurs out faces in home videos. I’ll have an automatic timer that blocks out my coffee machine after midday…

• 🔗 Link: Developer for OpenAI, Logan.GPT, highlights the importance of safety & collaboration for the future of ChatGPT. With regards to safety, OpenAI has stated its plans to remove access for those under 18 and cites GPT-4’s increased safety procedures—which makes it 82% less likely to respond to disallowed requests than GPT-3.5.

…………………………………………………………………………….

😮 1 Media Marvel

• 🔗 Link: LumaAI allows you to capture any 3D object and reproduce it with unmatched photorealistic qualities. Intricate details, reflections & lighting are their specialities—and their app store reviews are filled with positivity. My personal favourite: “This is scary good coming from a VFX artist”. I personally see it as a useful tool for super-imposing furniture in your new home (IKEA… if you’re reading this… yes. I’ll take royalties).

What was your favourite find in AI this week? Reply to this email and let me know!

— Alex