I haven’t posted in a while. But I have been seeing and reading about AI everywhere. From AI powered gadgets to the inevitable AI assistants and everything in-between (some of which I have in my 💩Cool Shit links below).

We’re still early in the AI software cycle. The question I’ve been asking myself is where will it go.

The AI Revolution isn’t comparable to the Mobile Revolution, as the latter was more a distribution revolution. Rather, AI is more comparable to the dawn of the internet. Or, more fundamentally, AI is an even larger-scale technology shift—it’s the dawn of a new discrete revolution that’s built not around computers acting like calculators, but computers acting like the human brain.

The Mobile Revolution vs. The AI Revolution (Digital Native / Rex Woodbury)

Distinguishing AI from the mobile revolution feels right. But calling it a larger technological revolution feels slightly premature. But it might be right.

Generative AI today

Generative AI right now feels like an evolution, not a revolution for two reasons. 1) We’re still using our devices in largely similar ways, and 2) it’s still all about processing data.

What’s most different about generative AI is how it’s processing data. Instead of accessing structured data to give a single defined answers it provides probabilistic ones. LLMs in particular have just been very good at doing the probabilistic thing.

There’s clearly a use case for LLMs to help create stuff - words, code, slides, whatever. But a large part of what we do digitally is still accessing specific information. We’re just starting to see how LLMs can be used to do this. Sure, the UI and how we search for answers might be a little different, but it’s still typing something into a website or app. Ben Thompson puts it well (emphasis mine):

ChatGPT’s plugin architecture gave hallucinating creative LLMs access to determinative computers to ascertain truth, not dissimilar to the way a creative being like you or I might use a calculator to solve a math problem. In other words, the LLM is the interface to the source of truth, not the source of truth itself.

ChatGPT Enterprise, Connectors and Small Businesses, Nvidia Competitors (Stratechery / Ben Thompson)

What’s unsolved about the future

I don’t believe the current use cases of gen AI will necessarily be the final blueprint of the future. That’s why calling it a revolution could still be the right call.

We can look to two factors here that might give some signal for how generative AI will change our lives.

1. The unsolved problems or limitations with LLMs:

   Open challenges in LLM research (Chip Huyen)

2. The unproven economic model of generative AI:

   AI is not a magical economic engine; it works brilliantly in some use cases (such as selling ads, helping coders write code faster, playing classic board games) but in many others it simply isn’t reliable enough (e.g., truly autonomous driverless cars, medical diagnosis, ChatGPT-style search).
   

   The astronomical valuations for Generative AI companies might be justified, but might well not be. Thus far, the valuations seem to be predicated on hopes and dreams, without really factoring in the serious engineering risks.

   What exactly are the economics of AI? (Marcus on AI / Gary Marcus)

What it all means

For the short-medium term, if hallucinations remain the more difficult technical challenge to solve, LLMs will likely continue as interfaces to sources of truth for some time. AI will continue to show up in our lives as features inside software or hardware products. We’ll see snippets of an LLM-centric digital future, like what Casey Newton describes after trying out Raycast:

To use Raycast is to get a glimpse of life after the web, or at least the web as we know it. It offers the answer you were looking for without you having to so much as open a browser. You summon the collective knowledge of the world — collective knowledge that was often obtained by these chatbot makers under dubious pretenses — and you return to your work.

How to use future AI interfaces today (Platformer / Casey Newton)

But the economic barrier today means we’ll likely see the 3 big players - Microsoft + OpenAI, Amazon + Anthropic, and Google - with the greatest advantage. They can deliver AI products at scale and any significant shifts in how we interact with computers will be locked within their respective ecosystems. At least while GPU demand remains higher than supply.

For the longer term, it will likely continue being a series of smaller shifts that don’t feel monumental in the moment. We’re not going to stop using our phones or stop searching Google overnight. What is certain is generative AI will increasingly become common in digital products. But how revolutionary it will be is still up in the air.

💩 Cool shit

AI Simulations - This is really fun, and an interesting peek at how LLMs could be used to simulate scenarios. I tried the FBI Hostage Negotiation.

Audio Atlas - A natural language music search engine. This feels like a really practical application of LLMs.

Wikipedia search-by-vibes - A similar type of product as above, framed more about the vibes.

3D Enigma - An interactive visualization of an enigma machine.

Open LLM Leaderboard - A site evaluating and ranking LLMs.

OpinionGPT - A GPT model trained to be biased. It’s a fascinating look at how bias impacts AI chatbots.

British Seaside Simulator - This is exactly what it says it is.

Artificial client - Receive endless feedback on your work from this AI client.

Roggle - A reverse Google. Get a list of search results and guess the search result.

Share this email with a friend because it’s your Last step for removing your E-mail from our list, Request Number: UNS-81199767