Sometimes to look like a genius you have to look incompetent for a bit.
In a place operating with SCRUM you're often expected to be shipping things every sprint, and sometimes with a large and important piece of work the experimentation, and the building of the thing, doesn't fit in a sprint, and if you tried to ship incrementally you'd just be shipping junk.
So you're going to end up giving some standup updates that do not inspire confidence for a bit. "I tried this thing, didn't work out", "still working on X, tried another thing", etc.
It's possible at the end of it you figure it out, and the result justifies the time investment.
The spiral is friction with a system that disincentivizes taking on ambitious goals.
I wouldn't just blame the engineer for not shipping smaller things, being less ambitious, talking more instead of working, I'd blame the company culture for disincentivizing deep work and ambitious goals.
Obviously there's a limit to this. You do need to ship something useful eventually. It's possible you spend extra time and still fail to accomplish the thing you wanted. It might be damaging to your reputation. You win some you lose some.
With AI there's less time between having the idea to do something or try something, and it being done or ruled out, so I see this being less of an issue now. The main barrier is knowing what to do and describing it.
Another interesting discrepancy is that people think current GPUs are maybe capable of running AGI but they still can barely manage photorealistic rendering of a single room in realtime, or simulate something like a shirt thrown into a pile of laundry. They can generate a video of it based on millions of existing videos, but not do a real simulation of light and physics in realtime.
An AGI doesn't need that level of detail to do most tasks effectively. To use the OP's example, a cat catching a bug out of the air does not need to run a fluid dynamics simulation of airflow over the bug's wings to be able to catch it. A cheap approximation of the flight path is sufficient. Perhaps some physical tasks will need that level of detail but many will not.
Is the human brain capable of doing real simulations of light and physics in realtime? Or does it hallucinate the details and represent some low-resolution mental image? You may be overestimating the capabilities of flesh-based neural networks and underestimating the capabilities of silicon-based neural networks.
That's a good point. It's just kind of surprising, there's been a lot of incremental progress towards rendering and simulation for a long time. If you'd asked 10 years ago if we'd have AGI or realistic realtime rendering and physics of a single room first I think most people would have answered the former.
The other challenge with real people is the emotional aspect. Easier and faster to tell ai the code it wrote is terrible and needs to be redone completely, or maybe you just end the session and use a new model. Trickier to do with a human.
Yes, it's that. You do probably end up wanting to store some "latest" denorm tables at some point, but it takes surprisingly long to reach that point, and isn't hard when you get there.
There are disadvantages to this, but it's a safe default. The alternative is possibly losing important data, finding out later you want historical records of things that are stored in kludgy separate tables, getting into more advanced locking situations, and having more complex DB migrations. Which I've had to pull teams out of many times.
I don’t have a lot of experience related to this so I’m just noting some things.
Some people in this thread don’t seem to think it’s that hard or overcomplicated.
When reading Designing Data-Intensive Applications my main takeaway was that event sourcing can make it easier to solve a lot of issues like performance, scaling, consistency, auditability, etc.
It would be interesting to look into what a low overhead way of implementing CRUD with event sourcing in Postgres would look like, then decide if it’s too complex.
Try making Hackernews lite. Users can post, comment on posts, vote/unvote posts, and delete their own posts and comments. CRUD tables might be post, comment, maybe vote. Event tables might be create_post, delete_post, create_comment, delete_comment, vote, unvote.
a. Opposed to someone taking my money against my will and the law just because they want to, “for a better life”.
b. Not opposed to someone taking my money in exchange for goods or services I want.
a. Opposed to someone moving into my country against my will and the law just because they want to, “for a better life”.
b. Not opposed to someone moving into my country because I married them and want them here.
There’s a whole spectrum between a and b, but I think most people are against a.
Legal pickpocketing is taxes you’re opposed to, or wages being garnished.
In theory people who say they’re only against illegal immigration are saying they completely agree with all policies regarding legal immigration, now and maybe into the future. Likely not what these people actually believe because while possible it would be a silly position. They’re probably just saying it to try to find some common ground with very pro immigration people. Likely a fools errand.
You've invented a whole new set of categories of pickpocketing that most people would disagree has anything to do with pickpocketing to justify a flawed argument.
To your last paragraph, no, they are not. There is no automatic implication that people will not change their minds depending on the situation. The argument is not that they can never change their minds, but that it is deceitful to pretend to only oppose illegal immigration in current circumstances and then go on to demonstrate they are also opposed to the currently legal immigration.
Breaking and entering is to immigrating as pickpocketing is to taking my stuff.
Sometimes immigration is legal. Sometimes taking my stuff is legal.
There’s often overlap between legality and what people support or oppose, but not always.
“I oppose something that’s already against the rules” is a simpler/easier argument, rather than saying “I oppose the against the rules thing, and some of the stuff that isn’t against the rules currently”. I think that's why people make this argument of "only against illegal immigration". I agree it could be deceitful, taking the easier argument. It’s also not a tactic that works long term though if it just results in the immigration they were against being made legal. A stronger anti immigration stance, legal or illegal, would be a better starting point.
To be fair there may also be some people who a couple years ago really did only oppose illegal immigration but were then exposed to new information, and changed their mind to oppose some or all legal immigration as well.
Aside from the disagreements between these parties, what about the precedent of running a non-profit, and then transferring all IP to a for profit when it’s convenient to do so?
I wonder if the government or taxpayers have a case to bring regarding that.
This case sets no precedent one way or the other on that question. The IP was transferred to the for-profit for fair value in 2019, and Musk never argued otherwise in this case.
The attorneys general of California and Delaware could challenge that 2019 IP transfer if they so wished on behalf of the public.
Transactions like that happen all the time and are not problematic if handled legally. Any of the interested parties could have sued over it, but none have.
The interested parties would be taxpayers. I think some groups are trying to look into it.
The issue is that they did R&D as a charity, donations to which are tax deductible, there may also be other benefits to being a charity during R&D but that’s a big one, then once the thing works, setup a for profit, sell ip at “fair value”, get some investment, then things are ready for business.
I read there’s no statute of limitations on a tax issue like this, so I guess it might be hanging over them indefinitely.
I’m not a big taxation and government fan, they’d probably just waste the money anyways. It does seem unfair OpenAI gets to use this loophole though, unless all companies can make their R&D investment tax deductible, and get any other benefits of this setup.
Because they got to participate in the early investment in the for profit entity.
> Early Angels (Reid Hoffman, Peter Thiel, and others): Approximately $10 million invested, current value $1.4 billion. That corresponds to a return of around 140x.
The non-profit received shares in the for-profit as a result of the transfer. Those shares are theoretically worth hundreds of billions.
If it had been a for-profit company contributing assets to another for-profit company, the transaction would not have had any different tax consequence.
Yeah. I wonder if the question "Did OpenAI steal a non-profit worth what is now maybe hundreds of billions of dollars?" will ever be answered. If not it will be one of the biggest heists in history.
Of course that question has been answered. They stole it.
Only people who are pretending to be confused are people for whom it's in their interest to be confused.
They created a non-profit with the intention of launching AI to make sure that it would stay in the public trust, and then once it became really valuable, they spirited it away into private hands.
It's just a fact. It's obvious from looking at the thing. The details are designed to confuse and trick you into thinking that what you can plainly see happening is OK somehow.
The gvt won't have much of a leg to stand on until open ai actually turns a profit, until then how are taxpayers cheated? And by then it will be a taxable for profit corporation anyway.
Can we just bring some reality to this conversation. The non-profit still exists, it owns a significant chunk of a now almost trillion dollar company, and the for-profit wouldn't be profitable if it weren't for Billions of dollars of investment from Microsoft.
There is no counter factual where OpenAI exists as a non-profit and still inexplicably gets handed billions of dollars of compute to train LLMs. The for-profit company is a different thing from the non-profit and it exists for perfectly understandable reasons and I'm unsure why anyone other than Elon Musk pretends this doesn't make sense.
What is your proposed counter factual where the non-profit entity retains all ownership of the venture, but somehow finds hundreds of billions of dollars to train LLMs?
An alternative chain of events might be one where OpenAI, Inc. only operated through income and donations, which would probably lead to them scaling more slowly. They could also take on loans.
Having thought about this whole thing more I believe that non-profits should not be a separate type of entity that gets special tax treatment. People have different ideas of what constitutes legitimate non-profit activity. It just opens things up for tax avoidance and scams.
The anti piracy stance would be making it illegal for a surgeon/person to copy the procedure another surgeon invented without payment. Copying does not require additional labor from the inventor.
Obviously we would like the inventor of the procedure to be compensated, but is it worth depriving other people, or potentially them dying, to protect "intellectual property"?
Tesla was built on American R&D and manufacturing and is competitive in terms of cost and features. It’s probably easier to not innovate and go for almost guaranteed short term profit by relocating existing manufacturing processes though.
I don’t think there was no alternative. I think it was an easier, more short term, and more selfish (in relation to fellow countrymen and descendants), choice.
And tesla replaced many workers with automation to stay competitive, and also has factories in europe and china.
I do think consumers have a responsibility here: for the last 30 years i've only had european cars. Even if they're a bit less advanced or a bit more expensive, it's a choice for a healthy local economy. Only the last car was a Tesla, as it seemed to be the best EV choice, but that will not happen again. Again, voting with my wallet, i will not sponsor the actions of a wanna be dictator.
This kind of specialization can be bad for factory workers as well. Doing the same motion over and over can lead to repetitive stress injury.
I think it's true for any worker that performing the same activity over and over can lead to issues that wouldn't occur if work was more varied. That isn't necessarily in a companies interest though since for them it's easier to train people on one task, and they're easier to measure or replace.
The predictable and measurable thing for developers is thinking and cranking out features.
In a place operating with SCRUM you're often expected to be shipping things every sprint, and sometimes with a large and important piece of work the experimentation, and the building of the thing, doesn't fit in a sprint, and if you tried to ship incrementally you'd just be shipping junk.
So you're going to end up giving some standup updates that do not inspire confidence for a bit. "I tried this thing, didn't work out", "still working on X, tried another thing", etc.
It's possible at the end of it you figure it out, and the result justifies the time investment.
The spiral is friction with a system that disincentivizes taking on ambitious goals.
I wouldn't just blame the engineer for not shipping smaller things, being less ambitious, talking more instead of working, I'd blame the company culture for disincentivizing deep work and ambitious goals.
Obviously there's a limit to this. You do need to ship something useful eventually. It's possible you spend extra time and still fail to accomplish the thing you wanted. It might be damaging to your reputation. You win some you lose some.
With AI there's less time between having the idea to do something or try something, and it being done or ruled out, so I see this being less of an issue now. The main barrier is knowing what to do and describing it.
reply