And the whole AI craze is becoming nothing but a commodity business where all kinds of models are popping in and out, one better this update, the other better the next update etc. In short - they're basically indistinguishable for the average layman.
Commodity businesses are price chasers. That's the only thing to compete on when product offerings are similar enough.
AI valuations are not setup for this. AI Valuations are for 'winner takes all' implications. These are clearly now falling apart.
It doesn't really feel like AI for coding is commoditized atm.
As problematic as SWE-Bench is as a benchmark, the top commercial models are far better than anything else and it seems tough to see this as anything but a 3 horse race atm.
When you have more users you get more data to improve your models. The bet is that one company will be able to lock in to this and be at the top constantly.
I'm not saying this is what will happen, but people obviously bet a lot of money on that.
Lots of people know trans people (by which I mean people that have decided to take some visible action to transition) by now - the people in the boonies and flyover states have access to the same internet as everyone else, so a lot of people outside the most progressive cities have transitioned. You have to remember that trans had been at the front of the culture for 10 years (since Caitlyn Jenner) by the time of the 2024 election.
Crashing the economy? In the past year the S&P 500 rose 14%, unemployment is at 4.4%, and inflation is around 2.7%. There are many things to criticize Trump for but the economy has not actually crashed.
Trump ran on an explicit promise to bring down grocery prices on day 1.
Grocery prices have continued to climb.
Absolutely nothing he has done could remotely be said to be aimed at bringing them down.
He has also instituted massive attacks on the power of labor, and on the offices that report on things like the unemployment rate.
"The economy" is not just the stock market; unemployment numbers literally cannot be trusted coming from Trump's BLS; and an inflation of 2.7% is, in fact, fairly high (it's 35% higher than the "target" rate of 2%).
Clinical studies today are funded entirely by the pharma companies, keep that. Selling at cost of production would be something extra, for patients who want the drugs but aren't enrolled in an actual study. The company doesn't get solid data it can use for regulatory approval, so making them donate the drugs in that case seems excessive.
A downside would be that for drugs that don't cost much to produce, patients might be less willing to enroll in the studies, given the chance of getting a placebo. That could be handled by shutting down the informal access while the study is enrolled, for anyone who's eligible. I'm sure there are other wrinkles that would need to be considered too.
Put this way, this seems reasonable. Beyond cell therapy, I don't thin the cost of drug is the motivation for not making it more freely available. 'Misuse' leading to potential liability or unjustified bad outcomes, along with some regulatory burden seems like the issue.
> My layman's opinion is that I would happily pay a lot of money to have a robot help me around the house: fold my clothes, do the dishes, whatever dumb menial labor.
Then why don't you hire a helper for that?
You just said you'd pay a lot of money, so money doesn't seem to be an issue. What is then?
I can't speak for the other guy, but as a person who manages humans at work: I'd rather have a robot at home.
1) I live way, way out in the middle of nowhere.
2) Humans are fickle, late, emotional. They have requirements in their own life that conflict with the jobs I want them to do.
3) Taxes. I don't want to deal with this headache. 1099 my cleaner or whatever?
4) In my version, the costs of owning the robot are less than the costs of hiring humans. If that wasn't true, then I'd reconsider. I probably wouldn't buy one until the cost switched like that, unless maybe it was open-source or something.
Here's another way to think about it: Amazon is willing to pay workers to do the job, but they'd obviously rather have the robots do it. The robots work close to free, don't complain, and probably do a better job (at the jobs they're capable of). Why wouldn't they hire a human for that? A lot of the same reasons.
FWIW, I emailed auntanns.com to ask what a combination personal assistant and housekeeper would cost:
> Thank you for inquiring about our services. I'd love to discuss with you further regarding the person you are seeking. Personal assistants do not do housekeeping and housekeepers do not have the P.A skillset to pay bills and make appts etc unless they are an executive level housekeeper. Rates for executive housekeepers range between $60-$65/hr and a minimum of 20+ hours per week, plus PTO, paid sick days and many also seek a health stipend.
It costs approximately $200 for our house to be cleaned once (by humans). We do it about once a month because we don't feel like spending $200 weekly). It would be great to have it ~continuously cleaned but we the cost/benefit isn't there for having a full-time person.
ChatGPT, for all its flaws, does actually exist and definitely isn't just a remote-control-based illusion, and some people even pay for it.
Optimus, the only thing we can be sure is real is the hardware, which is the least interesting part. But even if they really are running just on software without remote control, the one and only thing they've shown in any public demo that would actually be impressive, was voice comprehension in a noisy environment.
reply