I have been using Copilot for over half a year or something, and I haven't found it to be useful in Word, and it is rarely useful in Outlook. I don't use Excel a lot, so I get a lot of value out of the formulae generation stuff when I have to use Excel.
There are several meetings that I need information from, but don't need to attend. Culture has been to take the summary from a meeting and dump it into the chat after, and if it misses anything add it below it. Really nice for action items and things like that.
Are those actually... good? I've only seen the Zoom ones, but those are _comically, unusably bad_.
Maybe it won't be good, but it's surprising to me it can't do it barebones. Considering LLMs do an okay at summarizing into markdown, maybe the issue is a lack of proper intermediate DSL to represent slides it can train on?
But autogenerating at least a scaffolding for a deck seems so obvious, it's pretty damning it's so bad 2 years later.
I'd love the R part of RAG, but Microsoft can't even do that it appears?
They push the "generate image based on prompt" as the main add for slides to make them not just bullets, but no real foundational slide creation is baked in.
Things may have changed since I demo'd it ~6 months ago.
Google Slides is just generated images and transform text as far as I've seen.
We now have so much information that the effort needed to take this step is so obnoxious that need an abstract organization system to exist on top of it.
I wonder if the scaling properties here just serve to highlight the deeper organizational problem that we still haven't solved properly yet.
Yet, here we are.
Right now, they can chuck out a terribly implemented LLM.
Also, why are we linking to some forum and not the articles?
I tried talking to him that it's still garbage in and garbage out. If the solution was never there to train on, LLMs aren't going to make them out of thin air. There's also the "language" part of LLMs, that he didn't get.
We get pulled into a workshop with Amazon, who also wouldn't give us free resources unless it had a GenAI angle. After all this effort, you know what we made?
A fucking chat bot.
A fucking chat bot, that's also near useless, because our customers are highly driven and only interested in mathematical predictions. Turns out, you have to actually solve the hard math problems first with many examples, before GenAI can even take a terrible attempt at it.
Paying for copilot did not happen.
they even broke the Save dialog in an attempt to push their onedrive garbage
what used to take 1 click now takes 3-4
Now I have to hit the '+', at the bottom of the window, wait for the dialog to fully load, then click 'Attach file', at the top of the window, wait for the dialog to load again, then click 'Upload from this device', again at the bottom of the window.
I sincerely HATE my coworkers for asking for attachments now.
No one at Microsoft is paying attention to the product as a whole. It seems like teams of middle managers all vying to get their dumb feature included in the product. It's an incomprehensible mess of poorly thought out misfeatures.
(note: desktop Excel 2016)
There's an "Export" section in the awkward full-window UI that they replaced the File menu with, but instead of exporting it's just another way around to add extra clicks (File -> Export -> Change File Type -> CSV -> Save As) and get to the same "Save As" dialog that makes the CSV file take over your window as the active file after you "export" it.
You can also get to that via Save As -> Browse and pick CSV in the "save as type" dropdown, which behaves how you expect Save As to work (where that copy should stay active in the window afterward, as opposed to Export where it shouldn't). But this spot is also stupid because now have to go File -> Save As -> Browse because somebody at Microsoft wanted to insert an extra page where they could default your location to OneDrive and make you click Browse to get a normal file browser.
I saw an article yesterday about how Microsoft is moving toward "user-centric design" and removing ads from Skype, but how about doing that in the products anybody gives a shit about?
Being able to install 20 year old software on a new computer and have it work is pretty nice.
Of course, some people may think subscription fees and cloud storage are features, but to me they are bugs...
PowerQuery and the Data Model have more in common with python than the Excel of old
Then the lack of financial return on investment from the initial exploratory (try everything) stages. Leading to a sizable pullback in spending across the entire ecosystem.
Then the doom & gloom (talking heads on TV; smarter-than-thou industry pundits); death of garbage start-ups (the great thinning); painful layoffs; stocks decline.
And last, given some time - very substantial, tangible value (productivity gains, new products) is realized across decades. Rinse & repeat.
My favorite is the Trough of Disillusionment!
Anecdotally, I've heard demand for OpenAI's enterprise tier is ridiculous from someone working within their GMT function. I guess you can say Microsoft did a great job of hedging their bets.
This says pretty clearly that what Microsoft is peddling isn't even remotely an AI. They used AI purely as marketing term and ignored the obvious market consequences that would arise from that.
I seriously think Big AI (Microsoft, Open AI and other players) somehow influenced this burying.
It's super-damaging to Microsoft, who have tooted O365 "AI" like crazy. Not to mention their investment in OpenAI. Continued criticism like this about their clumsy "AI" is probably going to cost them a few billion.
This feels unnecessarily conspiratorial. Occam's razor would imply that it's just that a lot of legit HN users are pretty emotionally invested in this stuff, and dislike seeing negative stuff about it. You can see this in the comments on these sorts of articles; there's definitely a cohort who have a quasi-religious zeal about this stuff.
This, combined with the otherwise non-specificity of the article gives me pause.
Is AI bunk? Sure! Even the latest-and-greatest models can't explain how to programmatically create a DNS record in Azure using the latest C# SDK, how to create a searchable/sortable React data-table with a remote data source, or how to achieve world peace (just to cite 2/3 topics everyone utterly failed at the last month when I tried).
Is this article bunk? Yeah, most likely -- no exec in any pharma company in their right mind would admit failure after only six mere months, and without any lawsuits...