Microsoft’s AI CEO, Mustafa Suleyman, has shared his opinion after recent pushback from users online that are becoming frustrated with Copilot and AI on Windows. In a post on X, Suleyman says he’s mind blown by the fact that people are unimpressed with the ability to talk fluently with an AI computer.
His post comes after Windows president Pavan Davuluri was recently met with major backlash from users online for posting about Windows evolving into an agentic OS. His post was so negatively received that he was forced to turn off replies, though Davuluri did later respond to reassure customers that the company was aware of the feedback.
LLMs are cool up to a point. It is pretty neat to see something imitate human speech so convincingly. But that’s it: neat. Like a phone shaped like a hamburger.
For me to be impressed beyond that, your thing needs to either do something that I can’t do better or do something i don’t want to do. I can write a pretty good grade 8 essay. I can summarize a paragraph. I can write a dirty limerick. I can complete my own sentence without
punctuationsuggestions.When I want to create something, I want to do it myself. Putting my ideas in order and finding the words to say something meaningful is the fun part. Even when I write a work email or a cover letter, I’m not really interested in help. Either I have something to say (don’t want help), or I rattle it off in one go (don’t need help).
Microsoft, I don’t owe you my attention or my money. Make something useful that doesn’t suck. Impress me.
If they hadn’t jumped the gun so badly and tainted the launch with crap results, Google would have been well positioned to do something profoundly useful.
If it could actually extract useful information with citations and pointers for next steps and work as an interactive search, that would actually be really really useful.
The whole “hallucinating health advice” and “being terrible” thing really set them back, even if they’ve improved.
Like you said, I don’t really need help creating. I do need help remembering things or finding information: that’s why I’m using a search engine in the first place.
At work, there’s a person who knows everything about the job. He regularly gets questions where the answer is just the correct way to find out for yourself.
That’s what I want. “Oh, you mean X? Try looking at YZ. Oh, you wanted X, but in G conditions. That’s over in FOO. It’s confusing because reasons written down here…”
Good point on bringing this up. It’s imitation. The moment it needs to make arbitrary decisions it starts committing mistakes due to lack of context, which it doesn’t have because it’s a fucking machine following IF scripts. It is useless in this sense, but AI creators thought they could circumvent this by programming the AI say: “Oh! Sorry, you are correct, it is exactly what you said” after you give it the context only your human persona can grasp.
Had they called this letters or words calculator, it would be much more ethical than calling this AI. Hence why the bubble is bursting, lying to people about it’s capabilities. It is just a new calculator for phrases instead of numbers.
You’ve explained my feelings about llms perfectly. I have a lot of coworkers that use it constantly all day for work and are confused that I don’t. TBH I can see that it is saving them time on a lot of tasks, but I like that part of my job. It can’t do all the bad parts of our job (lab work) just the rewarding parts (reading, writing, statistics, figure design, coding).
I dislike AI (LLM) generated output for a myriad of reasons, but there has been one use case I’ve found useful, as much as I hate to admit it, and that is digesting large volumes of text and summarising it or identifying passages in that text that match my description of what I want to find - generally a better outcome than simple keyword searching, and I can look at the bit it identified, which is not AI content…it’s just pointing me to the right part.
And this makes me wonder why Google photos has such a terrible search function.
The problem with llms is that we’re basically going to a large library with books that are small and have a lot of info and we’re asking a calculator to do research for us which has a lot probability of accuracy.