Does anyone even want to try this? I'm a Windows user, M365 subscriber, VS Code user (when I'm not in Codex), had a Github Copilot subscription, etc... and a heavy AI user. Basically, the prime target for Microsoft AI. And every attempt Microsoft has made to put AI into a product has been unusable garbage, using the same models I find productive in other tools.
So many AI doubters I know hate using AI tools because their employer only gives them the Microsoft offerings. Which is a totally understandable reaction to those tools.
Totally agree. It is more of a rant than reasoning. Anyone who claims to know how this will end (good or bad) is extrapolating beyond their own knowledge and is vastly oversimplifying what is likely to be one of the most transformative technologies of the last 100 years.
In the last 50 years, only the creation of the Internet compares. Like with the Internet, nuclear, television, etc. it will probably be a mix of good and bad.
Very thorough write up. Clearly a lot of performance improvements across their stack. As if it wasn't detailed/long enough already, I would have loved to see some application-level benchmarks that gave a hint of the cumulative performance gains we should expect.
Same when sometimes you talk about AI with people and they say ChatGPT is absolutely horrible, gets everything wrong and never can keep track of more than 3-4 messages and so on.
Then it turns out they were on the free logged-out version, writing the most obtuse, confusing and contradictory prompts, and they're trying to do "real work" like this.
We're all having very different experiences with AI right now, and when people talk to each other about it, we're not nearly specifying what environment, tools, weights, etc are used, for these discussions to actually be valuable, because people are comparing apples to oranges everywhere.
Who should we believe? People who say AI scaling makes it a more dangerous threat than thousands of nuclear warheads? Or the ones who, whenever a new model comes out say "AI has already peaked. Not any better than Opus 4"?
Both parties are convinced that their take is so blatantly obvious as to not require justification. It feels like the only thing these kinds of takes justify is the point-of-view that nobody knows how this is going to play out.
Based on publicly available evidence, we should believe the former because AI models have made continuous advances that have been measured. The idea that the continuous advances will stop currently has no evidential support. I'm not saying it's not a possibility, just that it seems unsupported conjecture right now.
By the way, there seems to be a new form of AI skepticism emerging in the US that comes from general opposition to data centers, and in my experience the AI skeptic part of it is wholly irrational. I've met people online who suggested, without providing any evidence, that AI is useless and nobody wants it. That's a very implausible take.
As a technologist, I can use my imagination. There are incentives in both directions, but I am less afraid of being overly cautious and preparing for the potential harms.
AGI doesn't mean it'll be top 5% in everything. Half of people are below average intelligence. Most people couldn't profitably run a business. We could get to AGI and still have something that is as dumb as the average person on many things.
>Artificial general intelligence (AGI) is a hypothetical type of artificial intelligence that matches or surpasses human capabilities across virtually all cognitive tasks.
One problem in your analogy, for McDonalds, the franchisee is the one who actually operates the business.
For the Cybercab, Tesla is the one operating it via "Full Self Driving", their app, and the Tesla rideshare network. As the owner, the only thing you do is own it, insure it, have a place to park it, and take on the risk of what happens if the vehicle registered to you is in an accident. Clearly, a pretty one-sided assignment of risk.
Actually, while I still like your point there really is still a lot of operational work.
You need to clean, inspect, repair, insure, secure and charge the cars. To do so efficiently you will need to custom develop premises full of chargers and efficient charging and cleaning infrastructure.
That absolutely is operationally intense. Premises, permits, construction and then significant operations.
Maybe? It's how McDonald's runs things too. They make the food centrally and ship it to franchises. They take on brand risk, food logistics at scale and so on. Franchisees deal with local demand, operations, hiring, local laws and so on.
It's not an exact 1:1, but that seems to be the model here. Tesla does the things that can scale (FSD, manufacturing, etc.) and operators do the things that can't.
I think a more apt analogy is iPhone. You pay Apple for a brand new one, they make more money when you make in-app purchases.
Another really important issue is depreciation. You own the car for a few year and eat a cost of depreciation, and then figure out how to dig yourself out of the hole the way Hertz did. Similar to iPhone- you use one for 2+ years and then get a new version- you're also responsible if the phone breaks.
They could, but running local operations isn't his forte, preference or interest, it seems. Even if he had a massive fleet of these, he'd have to navigate local laws, demand, operations and logistics.
His revealed preference is to run things centrally and at massive scales.
> They could, but running local operations isn't his forte, preference or interest, it seems. Even if he had a massive fleet of these, he'd have to navigate local laws, demand, operations and logistics.
reply