If you spend anytime on Twitter you’ve probably seen the article Something Is Happening and the polarized reaction to it. For Price Per Token’s newsletter this week I decided to give you just what you need: another take on it.
For those who haven’t read it, the article gives an ominous, somewhat alarmist view on what AI’s advancements mean for the future of the economy: most knowledge work done on a screen will be handled by AI and most of the jobs doing that work now will slowly disappear.
The author, Matt Shumer, mentions several times that he even thinks that the new models now have “taste.” A word I see thrown out constantly, and in my opinion now a loosely defined cliche.
It is true, AI is advancing at a rapid pace, but we’ve all known and expected the exponential improvement for sometime, so, for me it is hard to be surprised at the new models. Maybe that’s because I run an LLM benchmarking site, I’m not sure. But that probably means I am not the target audience. Many of you are probably not either.
If you step out of the tech bubble it can feel like looking back 5 years. Many people’s day-to-day jobs have not changed much and still use AI casually- maybe asking ChatGPT to do basic research or to help write an email. Not take over major parts of their job like software engineers do now. So, for them, this article could be useful because a lot of jobs will look different.
But I am not so sure the alarmist view makes sense. The article mainly focuses on AI’s increasing capabilities and not around people and company’s incentives which are the driving factor behind adoption.
He mentions legal work, saying that AI can now “read contracts, summarize case law, draft briefs, and do legal research at a level that rivals junior associates.” First of all, if you ask any attorney what they think of the AI tools in their industry, like Harvey, you will find most are disappointed. So it seems that the tools they have available are not that capable, at least yet. Second, law firms make most of their revenue based on billable hours. The longer it takes them to do something the more they make- they are not actually incentivized to do their work faster. AI becoming widespread in the legal market would necessitate major changes in this business model, which has been around for centuries. That shift is probably the real bottleneck around this, not the models improving.
The article also ignores the concept of accountability. Would clients want to work with a law firm mainly run by AI agents to handle cases that could mean the life and death of their business? I’m guessing they would be more comfortable knowing a human has reviewed all the legal work. That is not to say AI will not help law firms, it definitely will. But I don’t think we should expect junior associates to disappear just as we shouldn’t entry level jobs in many industries to disappear.
In my optimistic view I see AI becoming a value multiplier than replacer. With the same number of people companies can cover much more ground in their industry. A law firm can do more thorough research on case and a software company can test more products. I am sure we will see plenty of companies choosing to cut costs but I would bet plenty look to expand and improve rather than stay the same with less. From an economic standpoint we would see the total pie expand, this is not a 0 sum game.