A new era of chips

Plus: The future of weather forecasting

Happy Friday! Before we get into this week’s newsletter, I wanted to take a moment to acknowledge some major news that just rattled the AI world: Sam Altman’s departure from OpenAI. I had the chance to initially process this news, and I’ve shared my thoughts below before getting into our usual top stories of the week. If you have any thoughts on this news or predictions for the future of OpenAI, feel free to hit reply and let me know.

Let’s dive into the biggest AI headlines of this week.

Breaking: Sam Altman Fired as CEO of OpenAI

OpenAI

Today, OpenAI made an announcement that sent shockwaves through the AI community: Sam Altman has been pushed out of his CEO position by OpenAI’s board of directors.

  • “Mr. Altman’s departure follows a deliberative review process by the board, which concluded that he was not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities,” the company said. “The board no longer has confidence in his ability to continue leading OpenAI.”

  • Chief technology officer Mira Murati will be the interim CEO, effective immediately.

Some personal thoughts: This news just blew my mind. Personally, I thought Sam Altman was an excellent ambassador for tech and had a very balanced approach to how he spoke about the advancements of AI. He was optimistic and excited about where it was all going, but always spoke cautiously of what could happen if things went wrong. He always came across as realistic and empathetic towards concerns over where AI is going.

Maybe that was part of the problem. Maybe OpenAI wanted someone in their leadership that never talks about the potential negatives. Either way, as someone who is an excellent ambassador for large language model tech, I'm sure he will land at another AI company and pave the way for them. I wouldn't be shocked to hear that Sam lands at Microsoft, Anthropic, or even somewhere as crazy as X.

Let There Be Chips

NVIDIA

While AI developers are still scrambling to get their hands on powerful—and scarce—GPUs, two big players just revealed their newest chips. 

NVIDIA, whose H100 chips power many of today’s largest AI models (including OpenAI’s GPT-4), just announced its new H200 chip—and it’s a huge performance leap.

The details:

  • The NVIDIA H200 is the first GPU to offer HBM3e memory—141GB of memory that stores nearly double the amount of data and transfers that data 2.4x faster than its predecessor. 

  • This increased capacity and bandwidth will allow H200 to train and run larger and more complex (therefore, more accurate and efficient) AI models, fueling the acceleration of generative AI and LLMs.

  • Starting in Q2 of 2024, the H200 will be deployed by an exclusive group of cloud service providers (including Amazon Web Services, Microsoft Azure, and Google Cloud).

Microsoft introduced its very first custom silicon chips, the Azure Maia 100 and Cobalt 100, designed to power its Azure data centers starting in 2024.

The details:

  • Both chips were specifically designed for Microsoft’s Azure cloud services, optimizing for performance, efficiency, and scalability.

  • They will be used to power some of Microsoft’s largest AI workloads on Azure, including OpenAI’s workloads as the company’s exclusive cloud partner.

  • Unlike NVIDIA, Microsoft doesn’t plan on selling its chips to other companies—yet. The company is still evaluating the market demand.

Big picture: As the backbone of the AI industry, chip development is bound to accelerate rapidly in the coming months—especially with increased demand for ever-powerful GPUs. Meanwhile, Microsoft’s update points to a growing trend: More companies are now developing their own chips in response to the ongoing chip shortage.

Today’s Future Tools brought to you by:

Taplio is the all-in-one, AI-powered tool—trusted by 6,100 LinkedIn pros and creators—to grow your personal brand on LinkedIn.

Have you been struggling to create high-performing content or engage with the right people? Or lack the time and focus to get results? 

Taplio helps tackle your LinkedIn challenges with its advanced features: AI-powered content, advanced scheduling, engagement, lead generation, and so much more.

🗞️ Other AI News & Articles

Better Weather (Predictions) Ahead

Google Deepmind

In today’s changing climate, it has become harder than ever to forecast extreme weather. But with great challenges come great AI solutions—enter GraphCast, Google Deepmind’s newest AI-powered weather forecasting system.

What’s the buzz? This week, Google Deepmind unveiled GraphCast, an ML model that consistently outperforms the standard weather prediction model in both speed and accuracy. If my own forecasts are accurate, GraphCast might just be the future of weather. 

Key points:

  • Trained on four decades of historical weather data, GraphCast uses deep learning—instead of complex physics equations—to form accurate weather predictions. It can also incorporate real-time data from satellites, radar and meteorological stations for up-to-the-minute forecasts.

  • GraphCast can generate 10-day forecasts in under a minute on a desktop computer. For comparison, conventional approaches require hours of computation on a supercomputer equipped with hundreds of machines.

  • The model was found to be 99.7% more accurate than the European Centre for Medium-Range Weather Forecasts’ (ECMRWF) gold-standard system. The ECMRWF is now experimenting with the tool themselves.

  • GraphCast appears to be particularly good at predicting extreme weather events—like the onset of extreme temperatures and the tracks of cyclones—with greater accuracy further into the future.

  • GraphCast’s model code is open source, so scientists and forecasters around the world can contribute to its development.

Big picture: With GraphCast leading the charge, we're looking at a new era in weather predictions: more accurate, more efficient, and potentially lifesaving in the face of impending natural disasters.

GPT-5 is Coming

FT montage/Bloomberg

While most of us are still busy catching our breath after DevDay’s flood of product announcements, OpenAI just dropped more news. Mere days after launching GPT-4 Turbo, OpenAI finally confirmed that it is working on GPT-5.

In a recent interview with the Financial Times, OpenAI’s former CEO Sam Altman revealed that his company is actively working on GPT-5, but needs to raise additional funding from Microsoft to make it a reality.

The insights: While Altman didn’t share any technical details, he implied that OpenAI is in the preparatory phase of GPT-5 development—meaning the actual training of the new model hasn’t yet started. Currently, the company is likely establishing training methodologies and curating datasets. 

Speaking of datasets, training GPT-5 requires a lot more data than its predecessor models. The data will come from public and proprietary sources, which lines up with OpenAI’s recent call to organizations to collaborate on private datasets. 

Moving forward: The details on GPT-5’s launch timeline and capabilities still remain unknown. What we do know is that the expectations for GPT-5 are sky-high given the improved capabilities unlocked by each iteration of GPT—and superintelligence is on the table.

Everything else:

  • OpenAI paused ChatGPT Plus sign-ups following an enormous spike in demand post-DevDay. 

  • Google Deepmind just launched its “most advanced music generation model” in partnership with YouTube.

  • Argentina’s presidential elections give a sneak peek into how AI deepfakes can taint political campaigns. 

  • NVIDIA is working on AI chips for China in an effort to avoid US chip restrictions. 

  • In its first acquisition since going public, Airbnb bought a secretive AI startup for some $200 million. 

  • OpenAI has a committee to determine when AGI is here—and it excludes Microsoft.

📺️ This Week’s Videos to Watch:

More important AI news: Dive deeper into this week’s hottest AI news stories (because yes, there are even more) in my latest YouTube video:

GPT-fever: We’re diving into OpenAI’s hot new GPTs feature. I walk you through how to build your own GPT (I made one that turns your face into a cat, of course) and some of the coolest examples in this new video:

Big stuff: Watch Stephen Wolfram outline his profound vision of computation and its role in the future of AI in this fascinating TED talk:

And that’s a wrap! I’ll be back in your inbox on Wednesday with five new and exciting AI tools for you to try out. Have a great weekend!

—Matt (FutureTools.io)

Interested in working together on a newsletter sponsorship, YouTube integration, or featured listing on the Future Tools website? I look out for the audience above all else, so I'm pretty picky—but get in touch here and let's see if we can make something happen.

If you haven’t already, be sure to follow me over on X. I like to share news and AI updates as they happen over there. And don't forget to check out all of the newest tools we've just added on Future Tools!

You rock! See ya next week. :)

P.S. This newsletter is 100% written by a human. Okay, maybe 96%.