Microsoft's Nadella is building an AI empire
wsj.com
wsj.com
https://support.microsoft.com/en-us/topic/gpt-builder-is-bei...
1. This is really cool! How amazing! We can do so much with this.
2. User activity drops off fairly quickly.
3. Oh they want money? Lets wait until the trial expires.
4. Not renewed.
They weren't realistic enough for sales (we tried), and all other use-cases seem a bit forced
But hey, they're growing and acquiring revenue (I think?) -- so who am I to question
Like the other trends we'll need for the marketing to die down and see what is actually real and useful. Once we hit that point the money is gone, no more crazy funding. It will also be interesting to see if we get helpful products that makes our lives better or if it's going to be damaging as we've seen with social media or relegated to the backrooms, like crypto-currency.
None of the technologies we've seen trending like this is good or bad on their own, but social media has been completely misused and rather than providing value and promoting human interaction it's used to track us and sell us stuff we don't need. Crypto-currency was, in hindsight, never going to be something that could be promoted to the average user, but is really useful as a backend technology. The current generation of LLMs have the possibility of been incredibly harmful, or extremely useful, but they aren't products in their own right.
Regarding the marketing hype, I think this video from der8auer at CompuTex was funny: https://www.youtube.com/watch?v=45CvTHmt_dI basically the AI label is just slapped on EVERY THING.
The only question is if AI has reached the "peak of inflated expectations" yet (in this case it's really high).
I still can't find any really useful examples anywhere. Unless we count gambling and illicit money transfers. All other associated activity appears like a rounding error compared to its non-crypto counterparts to me.
But at face value, if you could add a small seamless transaction cost to the web (idk, it takes 1c to read a blog post and 50c to post a blog comment), you'd basically eliminate almost every form of spam and re-align the incentives in the attention economy away from those of online advertisement. Ultimately it could just save the web from the mess it's become.
The banking and cards system isn't set up for that sort of workflow. Transaction costs are too high and you'd basically tie all your online activity to your real identity. So you need something new.
A distributed ledger would be almost required for future-proofing. There's an enormous juicy opportunity for rent-seeking if you assign a private entity the task of being the arbiter of every transaction of the new web; that needs to be designed out.
The one application where this isn't true is when regulatory risk prevents lawful verifiers -- i.e. illegal transactions. So, it's not coincidence -- proof-of-x will always be dominated by illegal user stories and desire paths as it is now, because that's where it's not strictly inferior.
Ransomware ? /s
https://www.msn.com/en-us/money/other/microsoft-s-nadella-is...
Text-only:
https://assets.msn.com/content/view/v2/Detail/en-in/BB1o7AZY
For a technology that's arguably going to be very core to a lot of businesses down the road, vendor lock-in looks to be an unmitigable risk.
With ageing demographics and increased education around the world, I think the next major productivity gains will be in robotics - replacing all the manual labour and care giving that needs doing in one way or another.
LLMs may be a stepping stone in that journey but hardware cycles don't move like software and hardware adoption is slow and painful. This is part of the reason why VC doesn't like hardware but loves software
Consider even the previous excitement around the metaverse, that seemed to fizzle due to lack of content. Well now presumably generative AI can (soon) create all the 3D content you like, which could open many new possibilities.
These will continue to be the bottleneck especially as we move to a deglobalised world, ageing and shrinking workforces in the developed world with population growth happening in EMs and FMs.
2. Company ranks their employees, lays off the lowest ranking percentage of people responsible for 10% of the productivity
3. Company reduces headcount
Rinse and repeat and then most people will be out of a job, unless stacked ranking is outlawed.
And it's not like those laid off can't start their own, disruptive ventures. As employees get more efficient, it'll be easier to run small, nimble companies that take on the big leagues. I'm quite optimistic about all of it.
AI will create a large power asymmetry which will inevitably result in large scale abuse with no limits.
In almost every economy where the elite does not need the support of the population to be wealthy and to exercise territorial control, the result is almost always a brutal dictatorship where the population is weakened and disarmed so they do not revolt.
The elites in the countries that suffer this rely on countries who are buying from them and if democracies made it illegal to do business with such countries many would not be sustainable.
And the curse can be avoided. Norway has successfully avoided the resource curse through creation of a sovereign wealth fund (Government Pension Fund Global) to manage surplus oil revenues but to accomplish this it had a long-standing tradition of democracy and rule of law ensuring government accountability.
AI could cause a problem similar to the resource curse. Because you could have a strong service economy while having 99% unemployment.
And who yields that power and can use it with little consequences will do it.
Why should elites allow the global population to grow indefinetely if it threatens their survival? Predictably the elites will use their advantage to compete against the masses for resources to ensure their continued survival.
That's a totally unrealistic take... on the scale of "communism". Not every person is a perfect communist, and not every person is a business man.
Now I know you may say that only some need to create disruptive ventures and hire others then, but that's also unrealistic due to power imbalances.
What will turn the wheel is how capitalism works right now: If too many people are unemployed and too few people can afford your mass market product, someone will need to give the unemployed a job or money so they can participate in the "free" market.
I'd say Copilot makes me 5% more productive, as in - MS CP bloat adds 5% to the stack of stuff I need to remove from clients' Windows systems. Most of that is because I have to open Edge and cleaning Edge is a particularly tedious process.
None of this is hyperbole, btw. Keeping up with endless new Windows bloat/nags/traps/guff is a regular part of my billable hours.
Or as Stafford Beer would say - The purpose of a system is what it does.
I had a conversation recently about this very thing. They were complaining that Salesforce is very hard to learn, even for simple tasks.
I asked why not use AI assistants? Because in practice they just aren’t capable of connecting the dots between documentation and configuration to complete the task at hand at the required level of customization. And that gap that exists between how things are stated to work and what’s actually required is where a lot of jobs are done and AI isn’t close to fixing. Which is why I still remain convinced that what’s being called AI is nothing more than a marketing tool to sell enterprise software and that jobs are going to be destroyed by it.
So, maybe it's actually helping offset the silicon valley investor war on compensation a little.
I think you need to get out of tech if you think every large enterprise runs on mac
I still believe that was the peak of their UX.
That might be true, but judging from desktop OS market share, what the competition is doing, sucks more.
MacOS is gaining ground, but because their hardware is legitimately superior. I imagine commodity laptops to be a dying breed in a few more years.
Gaming on a windows PC is the last bastion that macs tend not to compete in at all, and on that front there's SteamOS.
Microsoft signed their own OS death warrant when they switched from being backward-compatibility oriented way back in the Vista days. It might take another 2 decades for it all to play out, but from what i can see, it's one straight road down to obsoletion.
I haven't felt the need to use Windows for years now thanks to this.
It's not exactly there yet (especially that you need some technical knowledge to troubleshoot stuff). But I am getting similar performance playing games on Linux that I would be getting on Windows.
That being said, I would hate an always on AI always watching my actions and phoning back home to Microsoft.
At the same time I find categories like “god” and “evil” difficult when used in the context of corporates.
Maybe Apple can be described as more honest.
They sell you extremely expensive hardware so that they don’t need your data to be monetized.
I think that’s a fair deal.
But it will have consequences.
> The LLM bubble will be much bigger than the .com bubble from the 1990s. Its ultimate burst in 2026 will be also epic.
Like .com dark fiber, post-bubble dark GPU/TPU/NPUs would eventually find new use cases.