Copilot's crudeness has left Microsoft chasing Google, again A year ago it looked as if the world could be Microsoft's oyster. The software giant dominated the enterprise, was catching up to cloudy rivals, and then managed to purchase forty-nine percent of the for-profit subsidiary of ChatGPT creator OpenAI. Having secured a stake in the leading purveyor of generative AI, it started to …
"OpenAI carries a valuation of $157 billion. That makes Microsoft's stake worth nearly $80 billion."
for now until the inevitable happens and the AI bubble bursts and M$ share of OpenAI drop to 10 cents
MS, Google, AWS etc are all pushing AI like crazy knowing its a POS, but having invested eyewatering amounts of cash in this bubble their investors are demanding to see a return on their investments before it bursts and the cash disappears up the internets tail pipe
currently it is not going to happen,
having recently completed a AI qualification the main use cases are based around chatbots (which annoy the hell out of people) and knowledge bases, and businesses are not going to spend the money required to develop and baby sit / monitor them for the small financial savings
but the company W11 Outlook recently took to criticise my emails - all my emails - on accessibility grounds. Which is fair enough, I guess, except that it's complaining about my text being too low contrast. My black on white default text...
Copilot Meeting Summary
Action Items
All the time. It isn't just "Alice suggested we redesign this and people smiled, so we're going to", but if the people concluded that this made more sense than Bob's proposal to change the training documents and leave the code alone, then the notes will say that. If the redesign the code suggestion came up first and was discussed for three minutes, and then the documentation proposal was made and discussed for twenty, and then people said that they thought the training solution was less helpful than the code one, will Copilot properly understand that two thirds of the meeting might have been necessary to decide against something but does not need to be brought up in the next steps email? I don't know, but I wouldn't want to count on it.
AI in search is the exact opposite of what's needed. For years search Google and Bing (via DDG, of course) have been getting steadily worse at coughing up what commercial interests thing the user should have as opposed to what was asked for. Adding AI to the double-guessing isn't going to improve things for the user, just for the commercial interests.
As search results get fuzzier, they become ever less useful. At least to me. I'm usually searching by a very specific term. A part no, or a model number etc. Once its been through the AI fuzzer the useful results are overwhelmed by all the more common but not very close results. I search for YYZ -- because that's precisely what I need -- and get a bazillion results for XYZ, because that's what most commonly occurs out in the wild. Its fucking hopeless and getting noticeably worse by the day.
Hence why it will always be useless at spreadsheets or code suggestion or anything that requires accuracy and precision. Funny how we developers are being implored to use its random code suggestions but structural engineers and surgeons aren't... or at least I hope not..
A year (maybe two) down the road, after even more M$ money has been poured into Copilot development (and once it has been baked into corporate apps), it will be discontinued, in favour of a new, shiny, much better (and completely incompatible)..."whatever", using ChatGPT++
(See: Clippy, Bob, Cortana...et al)
Absolutely. After arguing that precise point recently on Reddit with what I later began to believe was a fanboy who believed Microsoft products were great because Microsoft said they were, I finally gave up. My points were mainly based on the fact that I had a pretty successful career fixing crappy Microsoft products. From my point of view, nobody will ever in a million years convince me that Microsoft products are anything but generally pretty crappy. OS, office suite, all of it.
"But...but..if they're so crappy, why are they industry standards?", he said. Are you kidding me? They're industry standards because the boss who makes those calls said "We're using Outlook for all our email needs" and that's that. Now the smaller companies that do business with them, to stay compliant, such as only receiving documents in Word format and nothing else, now it's a "standard."
Microsoft products are generally crap. There, I said it. I meant it, I have a ton of experience to back that opinion up, and I stand behind what I said. More to your point, their product life cycle appears to me to be "Look at our super new product! You have to have it!" Everybody buys it. Then they find out it doesn't work much better than the old version, maybe even worse, but Microsoft insists: "Maybe you're not doing it right. It really is a super product" they say. Finally the complaints mount until Microsoft can stand it no more. So they admit "Yeah, that did have its problems but guess what? We have a BETTER product now! You need it!" and the cycle repeats.
"...using ChatGPT++" And then "ChatGPT+++ 3.0", then "ChatGPT Extra", then "ChatGPT 2028" etc. 18,000 increments of the original whose current version is only marginally better than the original, maybe worse (as mentioned) with many features that users liked dropped from the product and features they hated "enhanced". Microsoft will tell you what you need, you don't tell them what you need. If you point out that you're the paying customer, I believe they internally pass that particular email around as an example of a hilarious notion.
That seems to be quite key for useful AI tech these days (vs stochastic parroting or randomly-thesaurus-augmented averaging). Brains and Machines has a nice (but long) recent interview with Carver Mead that mentions this (needle in the haystack) too, but from the neuromorphic perspective.
I wrote a hints and tips sheet on setting up WSL a while back and thought I'd have a play. I fed this to Notebook LLM and told it to produce a presentation. Now, it did have a limitation - it would produce the text of the presentation but not then chuck that into Google Slides. We're not quite there yet*. But I could also tell it to augment the simple instructions with benefits, drawbacks and use cases. Not bad - I have an extra training presentation for recruits to go with minimal effort.
Contrast this with the corporate Microsoft account... A lot of what Google Bundles costs nothing extra. Caveat: do not recommend feeding anything confidential or proprietary into any of Google's AI tools.
* Likewise, Gemini in Google Slides will produce a slide but not the whole presentation.
Microsoft has been offering a similar product for the past year in Azure AI Search, one-click and it can index your documents and data and give you a drop-in solution for an LLM chat bot for working with that data, you don't even have to upload the documents to their datacenters like you do with Google.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
