How to break the 'AI hype cycle' | Hacker News Reader

How to break the 'AI hype cycle'

mitsloan.mit.edu

42 pointsvedhsaka18 comments

This is a great article, but sadly there's just too much cognitive dissonance in the ecosystem for critical thinking and worse it's not rewarded by the algorithms, so articles like this end up making far smaller splashes than the hype cycles.

Almost exactly a year ago, Goldman Sachs produced a report showing the investment in AI is untenable. They showed up with the receipts. It was quite good. They also gave a fare shake to what's working and not working in the space, but most people didn't read it.

https://www.goldmansachs.com/insights/top-of-mind/gen-ai-too...

Thanks. I’ll ask AI to summarize it for me.

The Sloan school again uses the sycophantic conductor meme for the dear business leaders, many of whom are entirely uncultivated and have never touched a classical instrument.

Regarding "AI", at least this CTO is cautious, but also says: “Most jobs at this point can benefit from AI.”

That is of course incorrect. I'm waiting for a CTO who dismisses "AI" outright. That is probably too dangerous in the current climate.

AI has become the new IBM — nobody has ever been fired for recommending AI.

AI is shorthand for “we have no clue how to solve this problem, but AI can do it for us.”

I wish that that wasn’t the case. Generative AI is a very cool tech, not magic. And machine learning, the “other” AI, powers much of what we consider normal in our tech interactions.

> AI is shorthand for “we have no clue how to solve this problem, but AI can do it for us.”

I find this ironic, given that it is extremely difficult to have an AI solve a problem that you cannot, yourself, solve in some capacity.

Don't wait (just move away). Farm animals and circus animal don't change anything about how the farm or circus is run. They aren't selected for or trained to do so.

Triggers will come from outside. But thanks to "scale" even after that their response will be "too big to fail" protect our ass please.

That's how they survive after the Iraq/Afg fiasco, 2008 GFC. Wikileaks/Panama papers etc

Read like a sponsored content article for Akamai. For example, at Akamai, we use… blah blah

I don't think the current hype cycle ends until the next one begins.

Quantum computing is probably due for another round, or perhaps superconductivity.

I’m curious what can over take the AI hype cycle, I’m sure something, but in some sense AI feels like the hype cycle to end all hype cycles, but we’ve had AI hype cycles before.

Why does the human mind love hype so much? Everybody piling on without any critical thought.

To me it speaks to a fundamental weakness of the human psyche. We're all still mostly driven by herd behavior

Money? I could care less about hype, but maybe that’s because I don’t know how to monetize it

you're not wrong. the NFT thing would not die until we finally found another use for all these GPUs. AI took over a lot of crypto hype overnight, and while AI has slightly more use than NFTs something will need to come into the room and consume the oxygen

And to your point, I have seen more and more stories about quantum lately. Not a lot, but the noise floor is coming up a little bit

Akamai CTO Robert Blumofe offers four tips for business leaders striving to foster AI fluency by empowering employees with the right tools and best use cases.

useful insights but definitely a PR article.

The article mentions "Akamai" 7 times to just say:

> Don’t overinvest in large language models

> Don’t let LLM success cloud your judgment

> Explore the world of AI beyond LLMs

> Let employees experiment

a.k.a. nuffing.

Reject LLM, embrace traditional keyword SEO?

Was it hype? The hype was insane pre LLMs and then LLMs came out indicating the hype was accurate.

Now the hype is insane again.

Funny to see this on the front page right next to an announcement of a second model getting gold on IMO.

Watching anything slightly critical of AI be [flagged] on HN tells me AI is still overhyped.

tl;dr look outside llm if your use case supports it.

Reply on news.ycombinator.com