top of page
Writer's pictureDerek Corcoran

Eric Schmidt on the applications of AI to Warfare, Everyday Life, and Detecting Breathing Mice

SUMMARY
I always thought of Eric Schmidt as the older, geeky business guy that Google brought in to be the grown-up in the board room. But I'd never seen him speak so that opinion was based on, well, nothing. I caught Eric at the Den AI Summit in Denver; he was inspiring. The depth and breadth of his knowledge across technology, medicine, climate change, and genetics was mind-blowing.

I'm grateful to Eric for taking the time to attend the event and share his thoughts on the future of AI, applications to science, and things we need to be careful about. And yes, he talked about using AI to detect if mice are breathing when they're asleep.

Eric Schmidt at the Den AI Summit in Denver

Thanks to my friend Kang Ahn, I had the pleasure of attending the DenAI Summit – the country’s first city-led conference focused on AI and hearing former Google CEO and Chairman Eric Schmidt speak on the current and future applications of AI.


It was the first time I’d seen Eric speak. I always thought of Eric as the somewhat geeky business guy in the shadows of Google founders Sergei Brin and Larry Page. But my opinion was wrong. In the 45 minutes Eric was on stage being interviewed by Denver major Mike Johnston (an unexpectedly excellent moderator), Eric demonstrated the breadth and depth of his knowledge in a very modest and unassuming way. I walked away thinking “Wow – that guys is seriously smart and someone I’d love to have a beer with.”


Eric talked about applications for AI that ranged from everyday co-pilots to warfare, from climate change to genetics and yes, even detecting if a mouse is breathing when it’s asleep (apparently that’s hard to do – and I’m not sure why we care 😂).


Below are a few personal highlights from the session.


What’s next for AI?

It was an interesting place to start, but the first question posed to Eric by the session host was “What’s next for AI?” and he highlighted 2 capabilities that he believes we will see soon.


  1. The Infinite Context Window and Agents with Memory

    The idea being that AI Agents will understand the context of previous questions and build upon them. Question 1 > Question 2 > Question 3 etc. where subsequent questions understand the previous questions. They remember. And as such, the next answer is contextual. We’re already seeing this in solutions like Google Gemini. When you ask a question, it offers up suggested next paths to drill down into topics. But Eric gave the example of “I want to build a house, where should I start?” leading to “How do I get planning permission?” and “What kind of foundation would be best given where my land is” etc.

  2. Text to Action

    Eric talked about Agents that could perform tasks based on questions and their understanding / answers. Gemini for example can’t control my smartphone currently (maybe that’s a good thing!). But in the future these agents will go beyond Text Questions > Text Answers and begin performing tasks for us. Eric jokingly commented that Programmers never do what he asks. So Programming Agents will democratize programming skills and allow business people to ask the Agent to do something and the agent will write the code, execute it and produce the desired result.

 

We will have Agent Stores like we have App Stores

In many areas of technology and the digital revolution, we’ve seen specialization. The App Store is a great example. In December of 2023 there were approximately 3.8 million apps for iOS in the Apple App Store.


Regarding AI, Eric talked about Agents being something that will increasingly specialize, just like mobile apps. And companies will compete to get us, as consumers and businesses, to download their particular Agent to help us solve everyday problems. This will lead to Agent Stores, like App Stores.


And the real fun will begin as we start combining these Agents and nesting them. Schmidt commented on the mathematical complexity of Testing agents and Testing Nested Agents and how this becomes a very complex problem very quickly that could result in a flash crash. Hopefully, he was talking about systems crashing and not the stock market … but with everything connected, perhaps it's both.


Complex problems in genetics are benefiting from AI

Eric demonstrated his passion for science with his enthusiastic discussion of the application of AI to genetics. As someone who does not understand this, I thoroughly enjoyed his simple explanation that I will try to summarize here (hopefully I get this right).


Language is complex and involves sequencing words to make sentences.


Large Language Models basically understand what word should come next and can write incredibly well-structured sentences with ease.


The human genome is incredibly complex and made up of the DNA instructions found in a cell. The genome includes protein-coding DNA sequences and non-coding DNA sequences ... but they are sequences.


AI models similar to LLMs can understand what possible DNA sequences could come next. LLMs can help improve the accuracy and efficiency of DNA assembly, the process of reconstructing a complete genome from short DNA sequences. By analyzing the patterns and relationships between DNA sequences, LLMs can help identify and correct errors in genome assembly. This has implications for understanding diseases and developing drugs to treat those diseases. Billions of iterations can be cost-effectively tested (through modeling) to identify the shortlist of potential candidates worth developing and testing.


Eric talked about a new type of pain medication that is being developed – and what was interesting about the story is that AI models were used to generate thousands of potential synthetic options. Then the team selected 10 that were most different from traditional medication like Advil and Tylenol for evaluation … because as humans, we’re developing a type of immunity to Advil and Tylenol through overexposure. So we need something very different for it to be effective in the long term.


Mind blown!


Climate Change is complex and can be helped by AI

Eric talked about two areas of climate change that are benefiting from AI. Cloud modelling (apparently, it’s really complex and benefits from AI) and Fusion.

Schmidt said we need to have figured out Fusion as a way of generating energy by 2045 – after that we’re going to “cook the earth” if we have not figured out fusion as a viable energy source. And working towards viable ways of achieving safe and sustainable fusion are also benefiting from AI modelling. Building a mini sun (the sun is basically a massive fusion reactor) is complex. But that complexity is benefiting from the power of AI modelling.


He indicated that we need to use Gas, Wind and Solar as a bridge to Fusion over the next 20 years and then when we’re ready for Fusion, we move to that as our dominant energy source.

He also commented that energy distribution is another problem we also need to solve – with our current process it takes an average of 18 years to plan, design and develop a power line (planning approval, environmental impact assessment, purchasing the land, etc.) and that needs to speed up. He didn’t suggest AI would solve that problem – but did say we need to figure it out.

Sigh.


There are real risks in the application of AI to Warfare

On a darker note – Eric did touch on the importance of a Human In The Loop when it comes to warfare. Or what the defense department calls Effective Human Control. He commented on the Wargame / Terminator type situations where AI gains access to launch nuclear weapons and decides to launch with no human control.


But brought this back to the current day situations of how drones are being controlled by pilots at a desk drinking a coffee. We already have separation of humans from the front line. As we introduce AI into this mix, to help accelerate decision making – are we getting too far away from Effective Human Control. Can we trust something we’re only partially controlling to make decisions that could impact human casualties?


At the end of the day, Eric asked that we always have the ability to “unplug” the AI – to ensure we always maintain that Effective Human Control. And if the AI ever begins to do things we don’t understand, we should again “unplug” it.


The future needs young entrepreneurs, sustainable energy, and explainable AI

As we look to the future, Eric reminded the audience that the most valuable companies in the world were established by impossibly young entrepreneurs 20-30 years ago. Think Microsoft, Google, Apple, Tesla. With that in mind, he encouraged Denver to attract today’s young entrepreneurs and give them the environment in which to experiment, fail, and ultimately succeed. Build the education system and foster the innovator culture that we know ultimately builds great businesses and a better world.


He encouraged all parents in the audience to go out of their way to talk to their teenagers who won’t want to talk. And encourage them to learn Python.


He encouraged business owners to embrace diversity in the workplace as it leads to better businesses.


He said we need to think about “explainability” in AI and saw this coming in future algorithms.


And he asked that we focus on sustainable cost-efficient energy as we’re building technology that is energy intensive and the countries that can supply that energy will be the center of the data economy (along with the countries that can supply the people – but the people and the hardware don’t need to be co-located, just ‘connected’).


When asked what had set him on his personal path to success, he (somewhat controversially) said “I worked from an office 5 days a week and had mentors”.  As someone in their 50's, I was among the members of the audience nodding in agreement. Working from home is super convenient, but we're depriving the new entrants into the workforce of our experience.


 

Multiple legends throughout history have drawn a connection between Genius and Simplicity. One of my favorite quotes is from Albert Einstein: "If you can't explain it to a six-year-old, you don't understand it yourself". And whilst I don’t think I saw any 6 year olds at the Denver Convention Center leaving Eric Schmidt’s session – I did see over 1,000 people who appreciated the genius of Eric and his ability to make very complex topics easily understood. I think the vast majority of the people in attendance appreciated that they were in the presence of greatness. And I’d like to personally thank Eric for taking the time to visit beautiful Denver and educate and excite us all on the potential of AI.

11 views0 comments

Comments


bottom of page