AI News Brief for Friday, April 24, 2026
Why Experts Love AI and Everyone Else Isn't So Sure
Most of us encounter AI in its most inconsistent forms and may be forming negative impressions based on limited exposure. Why experts remain optimistic.
Even AI Has Its Off Days
Relatable tech-fail stories perfect for anyone who has ever been frustrated by technology.
Why Experts Love AI and Everyone Else Isn't So Sure
Free chatbots and customer service bots may be causing most of us to form a negative impression based on limited exposure. Understanding why experts remain optimistic helps put individual AI experiences in context.
If you've tried a free AI chatbot and walked away unimpressed, your experience is more common than you might think. It's not unusual to get responses that missed the point, confident answers that turned out to be wrong, or advice you wouldn't act on. You might be confused by tech experts who rave about AI like it's transforming everything, when your own experience has been decidedly mixed.
A new analysis in MIT Technology Review explains why both groups are right, and why they're so thoroughly talking past each other.
The gap comes down to which AI people are actually using. According to Stanford's 2026 AI Index, 73% of AI researchers and experts believe AI will have a positive impact on jobs. Only 23% of the general public agrees — a 50-point divide.
The MIT Technology Review reports explains that experts tend to use the most advanced, often expensive AI tools, many of which are significantly better at technical tasks like coding, math, and research. When you use those tools for what they're genuinely good at, the results can be striking.
However, when you use a free chatbot for something more open-ended, such as planning a trip, writing a personal email, or answering a medical question, you're likely to hit the "jagged frontier": the uneven edge where AI is capable one moment and surprisingly clumsy the next.
One AI researcher put it simply: "The degree to which you are awed by AI is perfectly correlated with how much you use AI to code." Or do other technical tasks (presumably).
That's useful to know. It means skepticism isn't a sign of being behind. Rather, it may be an accurate read on the version of AI you've encountered. And it means the experts' optimism isn't hype for its own sake; they're seeing something real, in a context most people don't share.
The honest picture is probably both: AI is better than a lot of people realize, and it's still not good at a lot of things that matter.
When you've used an AI tool, has it impressed you, let you down, or something in between?
https://www.technologyreview.com/2026/04/13/1135720/why-opinion-on-ai-is-so-divided/
Even AI Has Its Off Days
You've heard plenty about what AI can do. It writes essays, passes bar exams, beats grandmasters at chess. The capabilities are real. But anyone who has spent time with an AI tool has also had the other experience: the answer that was confidently, completely wrong. The recipe that made no sense. The response that missed the point so thoroughly it was almost impressive.
Reader's Digest recently rounded up eight genuine AI blunders into a humorous article. They're a good reminder that this technology, for all its strengths, is still has a few kinks to work out.
Among the highlighted oopsies: researcher Janelle Shane fed 30,000 cookbook recipes to an AI and asked it to create new ones. The results included a dish called "Beothurtreed Tuna Pie" with instructions to "surround with 1½ dozen heavy water" and another recipe calling for "1 hard cooked apple mayonnaise." The AI produced something that looked like a recipe, followed the structural conventions of a recipe, and yet produced nothing you'd want to eat.
That gap — between the surface appearance of competence and actual understanding — shows up across the list. AI systems that mistake context, misread images, or produce outputs that are structurally correct but substantively baffling.
None of this is meant as a reason to dismiss AI tools. It's more the opposite: understanding what they get wrong is part of using them well. A hammer that occasionally hits the wrong nail isn't useless — but you want to know which jobs it handles well and which ones call for something else.
And in the meantime, "Beothurtreed Tuna Pie" is at least a conversation starter.
What's the funniest or most baffling AI mistake you've encountered? We'd love to hear your stories.