Microsoft's Big AI Hire Can't Match OpenAI (newcomer.co) 25
An anonymous reader shares a report: At Microsoft's annual executive huddle last month, the company's chief financial officer, Amy Hood, put up a slide that charted the number of users for its Copilot consumer AI tool over the past year. It was essentially a flat line, showing around 20 million weekly users. On the same slide was another line showing ChatGPT's growth over the same period, arching ever upward toward 400 million weekly users.
OpenAI's iconic chatbot was soaring, while Microsoft's best hope for a mass-adoption AI tool was idling. It was a sobering chart for Microsoft's consumer AI team and the man who's been leading it for the past year, Mustafa Suleyman. Microsoft brought Suleyman aboard in March of 2024, along with much of the talent at his struggling AI startup Inflection, in return for a $650 million licensing fee that made Inflection's investors whole, and then some.
[...] Yet from the very start, people inside the company told me they were skeptical. Many outsiders have struggled to make an impact or even survive at Microsoft, a company that's full of lifers who cut their tech teeth in a different era. My skeptical sources noted Suleyman's previous run at a big company hadn't gone well, with Google stripping him of some management responsibilities following complaints of how he treated staff, the Wall Street Journal reported at the time. There was also much eye-rolling at the fact that Suleyman was given the title of CEO of Microsoft AI. That designation is typically reserved for the top executive at companies it acquires and lets operate semi-autonomously, such as LinkedIn or Github.
OpenAI's iconic chatbot was soaring, while Microsoft's best hope for a mass-adoption AI tool was idling. It was a sobering chart for Microsoft's consumer AI team and the man who's been leading it for the past year, Mustafa Suleyman. Microsoft brought Suleyman aboard in March of 2024, along with much of the talent at his struggling AI startup Inflection, in return for a $650 million licensing fee that made Inflection's investors whole, and then some.
[...] Yet from the very start, people inside the company told me they were skeptical. Many outsiders have struggled to make an impact or even survive at Microsoft, a company that's full of lifers who cut their tech teeth in a different era. My skeptical sources noted Suleyman's previous run at a big company hadn't gone well, with Google stripping him of some management responsibilities following complaints of how he treated staff, the Wall Street Journal reported at the time. There was also much eye-rolling at the fact that Suleyman was given the title of CEO of Microsoft AI. That designation is typically reserved for the top executive at companies it acquires and lets operate semi-autonomously, such as LinkedIn or Github.
eh (Score:2)
Not like DeepSeek isn't going to eat OpenAI's lunch eventually.
Re: (Score:2)
We'll see if OpenAI is successful at stopping DeepSeek from training their future models using ChatGPT.
Re: (Score:2)
"We'll see if OpenAI is successful at stopping DeepSeek from training their future models using ChatGPT."
How would they do that?
Re: (Score:2)
How would OpenAI stop them?
By being more careful about policing their API access.
How would DeepSeek train using ChatGPT?
The same way a smaller model is distilled from a larger model. You use the outputs of a larger model to train a smaller model to be smarter.
Re: (Score:2)
How would OpenAI stop any Chinese company from using ChatGPT to improve their product?
Re: (Score:2)
Training requires a lot of tokens to be generated. It would be a cat and mouse game, but at the end of the day- training usage is unlikely to have similar patters to "legitimate" usage.
Step 1 is knowing that it is happening [theverge.com], step 2 is finding ways to prevent it from happening.
Re: (Score:2)
It was looking like it, but the new image generation of ChatGPT seems to be unmatched and is for many users a killer feature.
I hope will will see equivalent open source models, but currently you need much more work for the same results.
On the upside allow the open source models with more work a larger variety of styles. Even though ChatGPT has some different styles, they all share some aspects making them look "like ChatGPT". I am not even sure if it is a bug or if they want to watermark their images that w
In other words (Score:2)
He failed at one company and was either able to bullshit his way into Microsoft, or they thought because he worked at Google he must have something.
Meanwhile, there are more qualified people out there who will never get the opportunity because they don't have the "right" background or contacts, or, hilariously enough, their resume/cv was rejected by the AI-enhanced HR filtering process.
Re: (Score:2)
Or, it might just be another case of Indians (dot) hiring Indians...isn't the CEO Nadella?
Github Copilot (Score:3)
Copilot users are ChatGPT users. They're just using it via Copilot. Or am I mistaken? I seem to be using GPT-4o according to Copilot.
We only see Copilot as a coding assistant tool. I can't imagine there are more than 20 million programmers using Copilot. I mean everyone is trying it out, seeing the limitations and what it can do. The CFO is probably trying to show that there's really no need to have Suleyman in that position. Which is correct.
Re: (Score:2)
Copilot users are ChatGPT users. They're just using it via Copilot.
There's no evidence for that, and quite a bit of evidence against it.
I seem to be using GPT-4o according to Copilot.
It's almost certainly lying to you ;)
Copilot scores *far far* below any OpenAI model released in the last couple of years.
It's worth noting that MS does, in fact, create their own models.
Phi, for example. Which is impressive for its size (it's tiny), but not at all impressive compared to larger models.
Re: (Score:2)
ChatGPT and Copilot are both systems using a model and not models. Both have access to OpenAI models as backend and ChatGPT wraps it in a nice webui and chat interface while Copilot integrates it with your IDE. It's like saying aren't web and mail the same, but meaning aren't they both using TCP/IP.
Microsoft "successes"?? (Score:2)
When was the last time M$ had a big win on a all-new product??
OpenAI is just the best... Period (Score:2)
Best image recognition.
Best at writing code.
Best at generative art.
Best at video now.
Best at every friking thing...
It is the benchmark to which all others are measured. And everyone knows it, even my mom. MS never had a chance... again.
Re: (Score:2)
Re: (Score:2)
Gemini 2.5 Pro is better at math and data analysis.
Otherwise, ya, o3+o4 are otherwise on top of the world.,br>
Re: OpenAI is just the best... Period (Score:1)
I feel 'best' is very subjective here. In my tests, the free Gemini 2.5 pro version through ai.dev and the Gemini app beats matches or beats $200 ChatGPT subscription on research, coding and integrations (at least with the apps I use). I don't do deep math so can't comment on that aspect.
Why anyone would pay $200 (or even $20) to OpenAI completely baffles me at this point TBH.
The $14 Gsuite enterprise plan that gives me all that OpenAI gives me for $200 and a lot more including 2 TB of free storage, own dom
Re: (Score:2)
In objective benchmarks, ChatGPT o3 and o4 outperform Gemini 2.5 Pro in just about everything except for math and reasoning.
o4 wipes the floor with Gemini in coding benchmarks.
I agree with you on the $200, though. That's fucking insane. As for who? My CEO is the only person I know.
Re: (Score:2)
Re: (Score:2)
Best image generation.
Best code? Nope, Claude is better.
Best image recognition: debatable. But sometimes you don't even want to use a LLM for ghat.
Best video? I think some of the Chinese models are better for some use-cases, but its hard to benchmark the overall performance of video generation.
Best at fricking? I think you'll need a uncensored model for that.
Stunning (Score:3)
It's quite stunning to hear that their usage volume has been so flat for so long. This is especially so when you consider how they have forced it into users faces, in Edge, in Windows,in Office, in GitHub...
20 million to ChatGPTs 400 million? Stunning.
Microsoft products... (Score:2)
...are increasingly untrustworthy, user-hostile garbage. People don't use their shovelware, because they don't trust it, and good thing too.
Hoping heads will roll in that company for trying to push the fad refuse on users, but knowing how much superficial groupthink rules MS, I'm not holding my breath.
PS: Burn Windows, use Linux. Kthxbai
They've throttled it pretty bad. (Score:2)
I wanted to get the instruction set (the old manual is online even) to a (60 year) old computer and it couldn't do it for "copyright reasons." Really?? What's more it's not allowed to say certain conteoversial words (not the 7 vulgar ones fwiw) and avoid them in a conversation. And it tends to be overly sweet. Not that I do t like flattery but it needs to be realistic. Also, it ends answers with questions to keep things going and it starts to feel smarny.
For these reasons and others it's not hard to see why
Windows 11 has made users finally realize that (Score:2)
it is an advertising platform focused on collecting personal information. Whether they realize it consciously or not, they are instinctively gravitating to "Not-Microsoft".
Naming Microsoft AI something else might have helped. There is a reason that GitHub and LinkedIn have kept their independent branding.
Given Microsoft's ability to ram their products down our throats, this is a pretty abject failure.
CEO of a department? (Score:3)
"There was also much eye-rolling at the fact that Suleyman was given the title of CEO of Microsoft AI."
Executive vice-president is the usual title for the head of a semi-autonomous department or special projects unit.