The Year in GenAI Part 1

It’s been an exciting year for GenAI - here are some of the highlights

As the year winds down, I wanted to do a post reflecting on the most exciting stories in AI this year. However, this ended up getting super long, so I decided to split that into two pieces. So, today you get 6-10, and then I’ll run 1-5 next week. Then, the following week I’m planning to do a post on some favorite (mostly non-AI content) I consumed this year. After that, I’m taking the next two weeks off, and I’ll probably return with a post on predictions for 2025.

As you can see, A lot has happened this year in AI, but also some things haven’t happened that we might have hoped would but that will come up next week.

So, here are 6-10 of my top 10 GenAI stories in 2024:

10. Self-Driving Cars are real

The number one tourist attraction in San Francisco right now is self-driving cars. They are everywhere, and tourists love them. Waymo is doing 100K rides per week in SF, LA, and Phoenix with Atlanta and Austin coming in 2025. Until you ride in one, this may seem abstract or exaggerated, but they really work. See the video below from a car I was in.

You can see that the car is able to successfully navigate city streets without error or driver assistance.¹ It wasn’t perfect, but across five rides, there were only two issues that required intervention and neither were safety issues. The cars also drove at a nice pace - the same speed as normal traffic.

I’ve always thought that self-driving cars were a technology that is perpetually 5 years away, but now they are here! If this technology becomes widespread, it will be a great help to the elderly, to commuters, and to bad drivers (like me). It will also pave the way for the jeansmobile.

9. Amazon Bets Big on AI + Robotics

One of the AI storylines that I’ve been following closely is the opportunity to use GenAI to greatly enhance what robots can do and how fast they can learn new skills. This is still a developing trend. There are lots of exciting pilots, but no GenAI robots seem to be deployed at scale yet.

The first proof point that the robots could be real was Amazon’s acquisition of Covariant. Covariant is a company that makes GenAI software for robots that makes them smarter. So, now you can tell a warehouse robot what to do, and it can listen and act on it. I’m assuming that Amazon bought Covariant because they saw promise in their technology and want to roll it out further in their warehouses.

I expect AI + robotics to be a big trend going forward and also expect to see lots more M&A activity in that space.

8. Three Mile Island is Coming Back

One of the phrases that I heard most often in 2023 from investors was “picks and shovels.” Everyone wanted to figure out how to make bets that enabled AI without having to pick a specific LLM as a winner. In 2024, one “picks and shovels” theme that has become popular is datacenters and power generation to support them. As LLMs attempt to train bigger and bigger models, the datacenter and power needs are projected to increase dramatically, and many companies and investors are trying to fill those needs.

One of the clearest statements of how big the power generation needs are likely to be is Microsoft’s announcement that they are reopening one of the reactors at Three Mile Island just to power GenAI models. As regular readers know, I’m quite interested in nuclear power,² and I’m quite comfortable seeing more of these plants open. I expect infrastructure to continue to be a big story going forward.

7. Survey shows investors considering GenAI impacts in diligence

In 2024, Generative AI has become a standard part of the investor checklist, especially in software and services deals. Our latest survey shows that 70% of investors surveyed have killed at least one deal due to concerns about GenAI risks, primarily about GenAI disrupting the company’s business model. Conversely, 71% mentioned underwriting GenAI opportunity on at least some deals, with 29% reporting that they do this on most or all deals.

This is another big shift from 2023. Investors are starting to look seriously at GenAI for potential investments both as a risk and an opportunity. I expect this trend to continue as there are more real GenAI proof points that emerge to justify underwriting.

6. Clusters of Effective Third-Party Tools

General purpose GenAI tools like ChatGPT are great (more on that next week), but I’ve been anticipating for a while that third-party tools will emerge that can do important jobs in a way that really makes it easy for companies.

This year we’ve seen progress in a couple of areas. Knowledge management, for example, is a place where GenAI is fully baked into the product. Horizontal platforms like Glean allow employees to chat with their data and generate new documents. For investors, new platforms like Sana, Rogo, and Hebbia have come to market that can deliver great performance for objectives like quickly interpreting a new presentation on a company or slicing and dicing a data room.

We’re also seeing some solutions in customer support from a few providers like Decagon that allow companies to automatically respond to customer queries even if they require the bot to take (simple) action.

It’s exciting to see real commercial grade GenAI solutions coming to market. It’s an important step to enable the business world to realize the full potential of the technology.

Hope you found this interesting. Please come back next week to see the events I’ve ranked 1-5.

1. You can also hear the calming music they put on by default to help the terrified passengers. As you can imagine, the first 5 minutes are scary and then it feels normal, and eventually you totally forget nobody is driving.

2. I actually went to nuclear engineering camp at NC State one summer in high school, which was incredibly interesting. Among other things, they let us do experiments with actual plasma, which were memorable.