Is Generative AI Hyped Up, or Is It The Real Deal?

Last week, Gartner placed AI on the “Peak of Inflated Expectations” in the hype cycle — a phase that many past innovations, including the internet, had gone through. And it’s not hard to see why…

• Money is flowing in the industry, more and more startups are pitching themselves as “AI” companies and AI products are being launched every day.
• 69% of the 1.5K companies surveyed by S&P Global say they’re working on at least one AI project.

Is the hype real? Gartner Analyst Arun Chandrasekaran says the productivity benefits are “nowhere close to what the vendors are claiming at this point,” per VentureBeat. And outside of chips and infrastructure powering AI, the revenue just isn’t there yet — which is “rumored to be in the hundreds of millions,” per Scientist and AI expert Gary Marcus.

Making the case that AI is overhyped

What worries Marcus the most is that the “generative AI economy — still based more on promise than actual commercial use — could see a massive, gut-wrenching correction.” And the problems with AI run deep, including its tendency to hallucinate, copyright concerns and another recent issue: declining quality of outputs.

• Research has shown that ChatGPT is becoming worse at basic math — with users pointing out a drop in the caliber of results.
• Model collapse is also a risk — where garbage outputs from the AI could be fed back into the model — worsening results over time.

For now, AI’s use is limited to tasks requiring low levels of accuracy. Per an Alteryx survey of 300 data leaders, 46% use it for content generation and 31% use it to code. The US Defense Department’s Craig Martell says he needs 99.999% accuracy before deploying the tech.

Forward-looking: Gartner claims that after AI drops off of the “Peak of Inflated Expectations,” it will cycle through the Disillusionment phase, climb the “Slope of Enlightenment” and coast into the “Plateau of Productivity” — where expectations become more realistic, knowledge spreads and mainstream adoption takes off. But it’s uncertain how long we’ll stay in each stage or whether we’ll even make it to the next.