Gary Marcus is having his moment in the sun, but the title tells all. “Be afraid” is just stripped down click-bait.
“Isn’t entirely candid” is the smoking gun of this article, which is frankly an absurd standard in today’s political environment, anti-tech media, and just the general hunger for an axe to grind 24/7 online. While I agree with Marcus’ critique on the vibe of the Senate hearing, it is a very strange pillar for proving Sam is “the man you should most fear on earth.”
Arguing that AGI hype/fears were caused by propaganda from Sam/OpenAI is recasting recent history. The entire media & cultural zeitgeist went nuts for a couple quarters when chatGPT went viral. Those fears have been around for a long time. A leap in perceived progress made them the Luddite topic of the day.
It doesn’t seem that hard to read the tea leaves here. If OpenAI could release GPT 5 like they released 3.5, they would. OpenAI’s greatest threat is that a genuine escalation in the publicly known capabilities of AI could exacerbate political push back and kill it. Involving government partners, in or adjacent, slows things down. Big surprise.
We can make regulation without claiming to be AI experts, ex: we can make deep fakes illegal where harm aligns with existing law, which many jurisdictions have already done. When AI regulation gets insane & dangerous is when people think they can invent levers to pull (like policing training data) that will supposedly ensure compliance. We all know what safety really means - a filter layer on interactions with the LLM. There is no fundamental solution to AI safety or regulation that OpenAI is abandoning here.
Marcus ends with his real starting line - he just doesn’t believe in GenAI. Fair enough. But building a massive ad hominem attack with a hand wavy “and some how Sam is still there” ignores the reality that OpenAI’s board now has some of the most trusted & trust-worthy people in tech on it. Sam’s level of accountability is higher, not lower.
The timing of this article is after the significant work OpenAI has done to re-affirm commitment to AI safety. Marcus here, and everywhere, deliberately ignores the mounting evidence that AI’s value is growing.
Which makes this entire campaign feel timed. Marcus’ is calling that the sky is falling right now. why so urgent? The investments have been made, the public markets are off the exuberant irrationalism, the major players are plowing generational levels of engineering into the quest.
Isn’t this what we should be proud of? The willingness to drive innovation forward without knowing for sure how the road will go?
Expecting giant technological leaps to deliver on massive value in a few quarters is for Wall Street. It has no place in science, innovation, or policy.