So you’re saying there are some upsides to AI?
Why do we talk about CEOs like anything they say is not complete bullshit?
you spent billions upon billions building data centers to host azure/openai, with no returns of value.
No you see, it was AI that made him do it. The CEO has no agency.
You’re the CEO. Do something about it.
CEOs don’t (usually) have surgical management skills. They operate on the body as a whole. Pills that affect everything. Some topical treatment to improve outside appearance, but usually that’s just makeup. Occasionally they open the body up and rearrange the insides to little effect. Ambitious CEOs will chop off limbs and attach other body parts hoping to grow something magical.
Oh you sweet summer child…
It’s gotten too big to stop. He knows he has to risk it all, or the other companies might do it first.
It’s like the nuclear arms race where they believed that nukes might burn the sky and kill us all.
Except, this time they might.
It would be the first time AI does something useful.
AI is very useful. LLMs are not the magic bullet that is being sold.
No–I’m sure at least one of the people it drove into a mental health crisis that ended in suicide was a fascist.
I read the article here is summary.
He is scared the people microsoft lays off will go work for their competitors.
He thinks if they lose “the AI race” they may be replaced by another company.
Microsoft still pledged to give $80 billion for their AI projects.
OpenAI wants to go for-profit but computing costs is not affordable for microsoft so they made a barely official agreement instead.
Workers suffered the most in this as the company has a culture where you may be sacked at any time due to a tweet from Musk.
My thoughts.
Microsoft won’t become irrelavent. Every single school and university go to unimaginable lengths to use their suite.
Highschools have a “microsoft word class” and colleges that pride themselves on their “academic integrity” give all students premium chatgpt accounts.
But they may do what AT&T did and go from a huge monopoly to just some random big company.
40% of their revenue is from Azure and servers. You’d think even if companies aren’t using MS branded AI products, they’ll still make a shitload of money providing the backbone for it.
It’s just like when they realized the lost the phone OS race and Ballmer left, they pivoted. They said, if we can’t run the OS, at least our software can be on all the phones. So, they push Office mobile, make their own launchers, put a bunch of investment into Intune and MDM management.
So you’re telling me there’s a use case for AI after all?
🤞
I mean, you’ve done an amazing job at destroying Xbox without it so I’m excited to see how effective a job your “ai” will do on the rest.
More CEO bullshit trying to convince us AI is going to be actually useful.
Not reading that he thinks AI is a good thing, just that MS may get out-competed on that front, and that’s a legitimate concern. That’s why all these tech giants are all-in on AI. They know damned well the bubble will pop. They’re gambling on being the last man standing.
I will say this - I suck at coding. I’ve spent years taking courses, learning basics over and over, only to be novice level at best. I can read code and KIND OF understand what it is doing, but writing it? Absolute dogshit.
LLMs have not raised my skill level in coding from a 15/100 to anything higher. It has, however, made it easier to generate code I want, can review, and use. It writes it at maybe 40/100, which is better than me. That’s the only thing it’s good at.
That’s not your fault! You aren’t bad at coding, and you aren’t bad at learning. Coding courses and tutorials are almost never assembled by anyone with any familiarity with pedagogy and they are famously bad at teaching. Tutorial hell is everyone’s experience.
If you actually want to learn, like any craft, time with your hands on the tools trying to make things is number 1. Literacy and familarity with jargon will help a lot as nothing is ever as useful as just reading the documentation and that is always dogshit.
Find a space you want to play in, like simple robots, maths puzzles, sound synthesis, simple games (godot is great for giving immediate visual feedback if you’re still trying to learn the basics), embedded systems/iot devices whatever. Then set yourself simple goals, first tweaking prior art and then building your own stuff. As you gain familarity work through a good textbook like The Art of Programming or https://en.wikipedia.org/wiki/Structure_and_Interpretation_of_Computer_Programs to get some ideas of the higher principles.
And stop doing tutorials, you will learn nothing. You need to study principles not practices, practices can only really be learned through independent practise.
I believe you can do it! and doing it yourself is really rewarding and playful. Don’t de-skill yourself and become dependent on a bunch of creepy fascist billionaires. Learning is a skill all skills can be improved to basic competence if you are sufficiently motivated.
Im a little bit the same. But I have more desire to keep learning the hard way than to use shortcuts. Using it as a learning tool is fine I guess. I just feel sick any time I use it and its not all that helpful vs finding something from a web search. Plus, its often wrong and teaches bad habits soo yeah.
Im always wanting to take the hard road though. I feel gross taking shortcuts on any task.
If you’re going to use it, use it to it’s fullest.
after it generates, aske it specifically what things do, ask it to go over the code and explain it’s decisions line by line.
ask it to teach you how to do what it just did.
It’s not the best code in the world, but for simple stuff it generates servicable code and can explain everything it all to you at a level to which you can understand.
best part, if you ask it and you still don’t understand you can keep asking it and it’ll work with you as long as it takes to get the point across.
i agree. microsoft still has a stupid amount of money printers spread out across multiple different parts of the tech industry. it seems unlikely that missing out on “AI” will be the death of the company.
I’ve been noticing weird stuff with Microsoft online products since they have been leaning into ai.
I don’t really know how to describe it. My old windows pc had a virus back in the early 2000s that slowly ate away at .dll and .exe files. It basically caused very random errors and noticeable weirdness for a few days until core system services eventually started dying. That’s what Microsoft online products feel like to use for me now. Random loops of confirmations and links that go nowhere. Even the payment system rarely seems to work.
His concern seems valid to me.
AI is destroying the company but not for the reason they’re thinking.
Alternate headline: AI FOMO CEO Satya Nadella doubles down on AI spending, fires profitable devs.
He doesn’t understand what LLM technology is and what they really can do and continues to buy the hype, so also continues firing staff (and whole departments) that generate income in order to employ ‘AI experts’ at exorbitant rates in the hope their magic AI will one day turn a huge profit - or at the very least to deny his competitors their skills.
The biggest issues Microsoft is facing at the moment are not AI: its EU (and wider) concerns about data sovereignty and surveillance driving businesses off their cloud, and their heavy-handed approach to spying on their users and forcing them into cloud services and apps in their ecosystem, which is driving users to alternative OSes.
Both are in the early stages and further loss could possibly be stemmed if the company strategy had a major overhaul in the other direction - but this whole direction has been Nadella’s doing, and a certain level of lost business is now unavoidable.
I hope they stay the path. The worst MS gets, the more interest and investment there will be in OSS alternatives.
Yeah, Nadella’s fear is that the company will be doomed if they don’t do AI hard enough, while any sane observer can see that his obsession with AI is leading them to throw hundreds of billions into a burning trash fire with zero gain to show for it. Microsoft still refuse to break out their AI earnings, which is proof enough that it’s making them close to zero profit (not counting the money OpenAI is throwing at them for compute). But everyone close to Nadella has been saying for a while that he’s genuinely monomaniacal about AI. He cannot see reason on this issue.
It’s Zuck and the metaverse all over again.
A cautionary tale that even the Zuck didn’t learn, given how much money he’s fruitlessly burning on bad AI and Ray Ban partnerships.
The only people actually making money off this whole thing are Nvidia, because if there’s one solid bet in a gold rush it’s selling picks and shovels.
You mean convince investors to GIVE THEM more money.
That would be OpenAI. Microsoft just borrows money against their share price when they need capitalization, they don’t have to beg for it.
God I hope so. Fuck microshaft.
This is less of a Fuck AI post, and more of a WTF is MSFT strategy reply.
OpenAI does not actually have that good of models, definitely poor value models, and a strategy depending on skynet contract from US military, which means having the absolute largest most trained most expensive model. MSFT has hedged its bets on decent alternative US model Antropic. But open source models, many from China, have genuine technical advances that make smaller (or larger) models that can be run within enterprise, or just you recent PC, viable competition.
MSFT cannot possibly make money by repackaging paid models from other providers that offer direct access. Just because copilot is in your OS and your office suite, does not mean there aren’t great alternatives that cost less, If MSFT ever made profit from its outsourced AI bunding, surely its own partners would make better offers to you for you to use them directly. LLM access still has massive free access options.
The US datacenter business is also massively energy constrained, with current administration only permitting expensive dead ender energy solutions, the moral of the story being datacenters being extra expensive. Local (smaller than openAI) LLMs in US will face similar electricity scarcity, but globally, local LLMs will prevail. MSFT being NSA/military/empire allies further prevents them from offering value to people/enterprise, and incapable of serving foreign markets.
