AI Godfathers + Muppet Corleone

Dr. Ryan Ries back again this week with some more big news from our field.

I had originally been thinking about all the advances in agents and vector databases that people are building and talking about, but the news stories just keep coming out this week, and I really feel like we need to talk about them.

First up, the world of AI just got its biggest nod of approval yet — the Nobel Prize in Physics.

News came out from Apple that LLMs are not able to do logic but rather just memorize common patterns.

Next, following in the footsteps of Microsoft, Google has plans to buy power from Kairos Power, utilizing 6 to 7 mini nuclear reactors.

Last, some tea has been spilled… OpenAI and Microsoft are at odds over infrastructure.

Let's look at each of these big news stories.

The AI Pioneers Who Made History

John Hopfield and Geoffrey Hinton, two trailblazers in artificial intelligence, have been awarded the 2024 Nobel Prize in Physics.

In 1982, John Hopfield created an associative memory that can store and reconstruct images and patterns in data. This laid the foundation for modern neural networks.

Known as the "godfather of AI," Geoffrey Hinton developed backpropagation in the 1980s, a technique crucial for training machines to "learn." His work is considered the birth of modern AI.

A Split Among the "Godfathers"

Here’s where it gets interesting. Both Hopfield and Hinton, while celebrating this achievement, are also sounding the alarm about AI’s potential risks.

But interestingly, not all AI pioneers share the same outlook on the field's progress.
Yann LeCun, who won the A.M. Turing Award (the highest prize in computer science) along with Hinton, has a bit of a different view.

He argues that today's AI is far less capable than many believe, and he even goes as far as to say it's "the beginning of a hint of a design for a system smarter than a house cat."

(Ouch, don’t show this to your house cat).

Apple showing that AI is memorizing patterns?

Apple researchers found that changing small details in math problems can confuse AI language models, suggesting AI relies on memorizing patterns versus understanding logic.

I wouldn’t say that this is news. Fundamentally, LLMs don’t have logic/reasoning capabilities.

LLMs are primarily designed for generating content through prediction rather than understanding logic.

To address this, researchers are exploring approaches like using "agents" to validate LLM outputs and developing new models, such as OpenAI's o1, that focus on step-by-step problem-solving.

While LLMs have impressive capabilities, their ability to reason logically is still evolving, and ongoing research is crucial for advancing their capabilities towards artificial general intelligence (AGI).

Where do we get the power to fuel AI?

Google has signed a major deal with Kairos Power to purchase energy from a fleet of small modular nuclear reactors (SMRs).

This is a big step towards Google's efforts to expand data centers with clean, reliable energy, especially as the demand for AI-driven services grows. The SMRs will be built in California and are expected to be operational by 2030.

While SMRs offer potential advantages like smaller size and faster construction, they also face challenges such as higher costs and regulatory hurdles.

The success of this initiative will depend on the ability of SMR technology to overcome these obstacles and deliver reliable, affordable energy.

Overall, this deal is a milestone in the development and adoption of SMR technology, and it could have a profound impact on the future of clean energy and the tech industry.

Trouble amongst other “Godfathers” in AI?

Now for the tea.

OpenAI and Microsoft are experiencing a shift in their partnership dynamics following OpenAI's recent $6.6 billion funding round.

OpenAI's leadership has expressed dissatisfaction with Microsoft's pace in providing essential server resources, leading OpenAI to consider alternative data center partnerships.

For example, their substantial agreement with Oracle to build a major AI data center in Texas.

This news shows us that OpenAI is looking for greater independence as competition within the AI industry intensifies.

Despite Microsoft's significant investment of over $13 billion and substantial profit rights, concerns about data center costs have made the company more cautious in its financial commitments.

My Thoughts

I am not even sure what to think about all the news this week.

It has been crazy to see all the stories coming out and the thoughts around when we might have the resources to create AGI.

As you know, I usually discuss how these models are here to help us and aren’t perfect, so all the discussions on AGI aren’t really surprising to me.

To be honest, right now, I don’t think it matters how close or far we are from AGI.

We’ve seen (and my team has built) tons of use cases for customers using AI for good.

At the same time, there are already situations where AI is harmful (take a look at last week’s Matrix and AI’s impacts on the environment).

Also, look at everything that is coming out about AI propaganda and people trying to understand what truth is, which at one time didn’t seem like it should be a question.

At the end of the day, whether you’re building AI solutions, selling AI solutions, or just here because you’re fascinated with all this technology can do, I think we all have a role to play here.

We need to continue to push the boundaries of AI while always keeping ethical considerations at the forefront.

Take the time to educate the public about both its potential and its risks. People shouldn’t be fearful of AI. But we should be educated about its risks.

What's your take on the Nobel news? Are you excited about the recognition? Concerned about the warnings? What do you think of OpenAI and Microsoft starting to question their partnership?

Until next time,

Ryan Ries

Now, time for this week’s AI-generated image and the prompt I used to create it.

"Generate an image of 3 muppets in the style of the movie The Godfather. These 3 characters are debating on their stance on AI's impact. You should see something in the image that shows what their intense discussion is about."

Sign up for Ryan's weekly newsletter to get early access to his content every Wednesday.