The AI news cycle has shifted. Again.
All eyes on Jacob Coxon, his resignation from Anthropic and his beliefs about frontier AI labs.
There are a lot of curious dynamics to consider. Here’s the framework I’m using to make sense of it.
1: Developing Your POV? Frontier Research is Where to Start. Not Where to End.
The gnarlier the researcher, the more weight I put behind their data driven, hands in the dirt and objective analysis.
On the frontier, they’re literally inside the building where the electricity gets made. That perspective is extremely useful but remember how important it is to ask the right questions.
Also, how important it is that you ask the right people…
*link to Clem Delangue 🤗's tweet
Reminds me of Benedict Evans's response to Matt Shumer’s “Something Big is Happening” back in February as well as the lack of attention the recursive AI dynamic received by the media at that time vs everything else.
*link to AIWA TOT #06 featuring Benedict Evans’ quote
2: What We Know vs What We Think. What They Saw vs What They Predict.
Those aren’t the same thing.
While it’s incredibly difficult to keep pace with exponential change, it might be even harder to predict the future. Regardless of which category of stakeholder you’re listening to, it’s critical to consider what they’re describing.
Are they describing objective data or making informed projections? Using the word “could” isn’t the same thing as “it did, therefore”...
How do we determine the likelihood that one person’s projections are more credible than another's? What’s an appropriate amount of downside risk protection vs enabling the extraordinary benefits / upside to be realized (i.e. curing cancer)?
How do humans understand what models are doing today vs in the future? Is there a risk that models will eventually do things that exceed human comprehension? Having a scoreboard vs not being able to see the scoreboard?
3: AI Twitter is Where You Can Distinguish Signal vs Noise. Fast.
AI Twitter is where signal vs noise can be found by looking at things through the collective lens of those closest to the action across a spectrum of perspectives.
Don’t take my word for it, here’s a list of 138 AI Twitter handles. The way they collectively interpret situations will get you to signal vs noise faster than any other method.
*Tweet link by Clem Delangue
Now, let’s run it on something at the heart of all of this noise.
Companies Regulating Themselves vs Companies Being Regulated. Regulating AI Here in the US vs Regulating it Globally.
Remember what happened on July 24. Jensen Huang published his “Open Weights and American AI Leadership” letter. His first tweet ever.
Within hours, Satya, Sundar, Zuck, Elon, Mira, Altman, Gurley, Jack and Brian Armstrong all answered. They lined up behind a policy position within a few hours.
Aligned incentives, genuine conviction or something else? From the outside, you can’t fully tell.
What you can tell is where it left the architecture.
Open + Closed. Not Open vs. Closed.
*link to AIWA #29
Dig in.
Humans + Machines. Never Humans vs. Machines.
