(2024-06-03) Procopio The Problem With GenAI Its Too Much Gen Being Sold As AI

Joe Procopio: The Problem With GenAI? It's Too Much 'Gen' Being Sold as 'AI' I’ve been neither a proponent nor an opponent of genAI, but I have pulled some of the threads on either side of the pro/con argument.

The problem — my problem — with generative AI isn’t the science. At all. It’s the application.

Back in 2010, I was the co-inventor of the first commercially available genAI platform, Automated Insights, which we sold to private equity in 2015 after generating over a billion machine-written articles covering everything from fantasy football to quarterly earnings reports and much more

To dumb my contributions down a bit, I hold a patent for the part that told the computers what to say, and the CEO holds another for the part that made the computers say it. I got out of the genAI arms race in 2018 and I didn’t look back

The aforementioned recent criticism was in response to last week’s post, in which I said that the corporate and venture investment worlds had completely over-indexed on AI: Thus, there is a ton, A TON, of venture capital and public company money sunk into some of the more … “non-genius” applications of the current flavor of AI

What did I get wrong? I didn’t state my “con” side strongly enough. Or my “pro” side.

AI Is Awesome!

when I talk about my “pro” side of genAI, I get equally called out. In fact, I’m getting railed right now for speculating on what I firmly believe genAI is going to do to today’s SaaS players.

Thus, I have three problems with today’s genAI arms race that I need to make clearer.

Problem No. 1: The Pursuit of the Lowest Common Denominator Applications

genAI is still way more gen than AI, but it’s being sold as AI, with these “headline big money use cases” that are not only obfuscating what genAI really is, but are doing so at the expense of the reputation of the actual science

And some of the stupider applications of genAI are also some of the more ethically empty.

Problem No. 2: The Death of Creativity

I talked about Automated Insights earlier. When we first developed the technology in 2010, no one understood it. And no one wanted it. See, almost all these teams already had at least one human writer covering them, and what those writers produced was often a far superior product to what we were offering. We immediately realized that genAI wasn’t a replacement for creativity, but a substitute for reports, charts, and infographics. Which, by the way, is part of my take on the threat to SaaS.

So problem No. 2 is the headline use cases themselves: genAI producing avatars to eliminate actors, speakers, writers, artists, technologists, advisers, and so on — devaluing the process of creativity and the value of experience

When I say that AI is going to be a failed attempt to technify labor, this is what I’m talking about.

Problem No. 3: It’s Still More Gen Than AI

AI that powers self-driving cars. It’s thinking AND acting. Self-driving cars are not operating at the same level of consistency and accuracy as genAI’s computer-generated images.

The “acting” part of genAI has evolved exponentially. The “thinking,” not so much

AI, true AI, is still either boring or dangerous. It’s boring because when it does work flawlessly, it is consistent and accurate only in limited applications in niche scenarios with few outliers. It’s dangerous when, like we’re seeing in self-driving cars, the operators try to remove too many limits and apply it to too many scenarios in which it faces an overabundance of outliers.

if we don’t change direction, and relatively quickly, we’re looking at evolving from the “chatbot and deepfake porn” stage of genAI to the “NFT and memecoin” stage of genAI.