mlinsey

8 ms·

61.
by mlinsey 2y ago
A tougher academic knowledge benchmark is great, but for something to be truly be worthy of the title "Humanity's Last Exam", I expect something more like: 1. Write a novel that wins the Pulitzer Prize. 2. Prove (or disprov
62.
by mlinsey 2y ago
There are a couple Twitter personalities that definitely fit this description. There is also a much bigger group of people that haven't really tried anything beyond GPT-3.5, which was the best you could get without paying a monthly sub
63.
by mlinsey 2y ago
People saw how much cheaper it was to train DeepSeek v3, and assumed this reduced NVidia's TAM. I think this doesn't make much sense. a) For inference, cheaper and faster compute will increase total inference spend, because the e
64.
by mlinsey 2y ago
I haven't seen any details, but isn't this a pretty straightforward way of doing it? The non-profit has had majority ownership of the for-profit subsidiary since 2019. The already-for-profit subsidiary has owned all the ChatGPT
65.
by mlinsey 2y ago
Because modern ML is a descendant of data science/stats work being done in systems like MATLAB and R. Numpy and scipy were great libraries, and Python was a high-level language that statisticians could use, but also a reasonable enough
66.
by mlinsey 2y ago
The flip side, is that they would not have been able to execute so well with Azure etc if the Windows org had too much of a say about pushing Windows as the OS of choice everywhere. Winning in a brand new space, especially one that might b
67.
by mlinsey 3y ago
The comfort, cost, and battery issues seem like must-solves before this is a true mass-market success. But they sound like the kind of thing that can be improved over time. As for the anti-social aspect...you'd have done very well for
68.
by mlinsey 3y ago
> when someone talks about codegen, it usually means AOP, bytecode manipulation, compiler plugins and the sorts. Not "AI tools" I have found the opposite to be true, starting this year. You can commiserate with people who are
69.
by mlinsey 3y ago
if Quiet doesn't catch on to provide that noise, would it be beneficial to flood Tor with LLM-generated conversatinos? We're really good at "cheaply generate near-infinite quantities of human-looking text" these days...
70.
by mlinsey 3y ago
"Why do we have not so many great individual players anymore but great teams? " IMO this hasn't changed as much as it seems from the outside. The brilliant individual leaders of the past owed most of their success to their te
71.
by mlinsey 3y ago
support in opinion polls, maybe. but actual union membership continues to fall and is at historic lows. Maybe if the current wave of strikes score some high-profile wins, that could start to reverse. https://www.bls.gov/opub
72.
by mlinsey 3y ago
We aren't talking about public spending on self-driving car infrastructure vs. bikes and mass-transit infrastructure. I would much rather fund the latter with public money. What we're talking about is erecting regulatory barriers
73.
by mlinsey 3y ago
Not sure what you mean, the movement to combat climate change is orders of magnitude bigger than the movement to combat AI risk - in terms of organizations dedicated to it, dollars donated, legislation passed, international treaties signed,
74.
by mlinsey 3y ago
In the scenario where the current AI boom takes us all the way to AGI in the next decade, IMO there is little downside. Risks are very large, OpenAI/Sam have expertise, and their novel corporate structure, while far from completely-rem
75.
by mlinsey 3y ago
Drawing a bright line between research vs products and considering only the former to be "innovation" is a way to become quickly irrelevant.
76.
by mlinsey 4y ago
> I’m seriously confused why anybody thinks anyone will even remotely care about any kind of pause is asked for. I can say that Google and Microsoft would definitely care, just as two specific instances of "anybody" that I have
77.
by mlinsey 4y ago
GPT4 is expensive enough to train that most organizations with that high of a compute budget also have compliance departments and will not flat-out ignore a law like that. State actors overseas and the indie hackers with open-source models
78.
by mlinsey 4y ago
I'm not convinced we understand qualia/feelings well enough to definitively say. The presence of eg phantom limb pain suggests pain can arise in the mind, independent of physical nerve signals. It would be a remarkable convergenc
79.
by mlinsey 4y ago
I think it's downvoting. It's one thing to see (what you think is) a really bad reply to you get some likes, it's another to see a negative number next to your comment, I feel much more compulsion to defend myself for the lat
80.
by mlinsey 4y ago
the "eggs in one basket" argument only makes sense if Martian colonization is itself just step 1 on a list of dozens of similarly-difficult leaps to get us to terraforming and interstellar colonization. It's a very exciting,
81.
by mlinsey 4y ago
Mobile apps or a web portal absolutely could have helped, simply by massively increasing the throughput of crew members being able to report that they didn't make it to their assigned location. Which could have stopped the cascading ef
82.
by mlinsey 4y ago
As far as I can tell, two groups are making plenty of money off of esports: 1 - The companies that make the game. Whether it's Riot and Blizzard selling slots in their leagues for eight figures, Valve using their annual tournament to
83.
by mlinsey 4y ago
This is why it is really just Harvard (likely Stanford, Yale, and Princeton too, but definitely Harvard) that can make a statement boycotting the list, just like a billionaire can get away with wearing a plastic $10 watch but mere millionai
84.
by mlinsey 4y ago
It's 2,000 shareholders, and you are allowed to go above the limit, it's just that if you do, then you have to make public financial statements and in general follow all the disclosure rules and reporting requirements of a public
85.
by mlinsey 5y ago
I think in the scenario where miners block the move to PoS (or any other much-more-scalable technology) is a scenario where nobody ever actually uses crypto for anything, and almost all coins go to $0 after the speculative bubble pops. To
86.
by mlinsey 5y ago
doing just a quick search in the public directory, I found 24 out of 394 companies are in crypto. That's a lot lower fraction than you'd expect given how much mindshare, hype, and attention crypto has now vs. other sectors. I am
87.
by mlinsey 5y ago
Your link supports the comment you are replying to; it says that after the reactor is shutdown, additional cooling is needed for some time, and that eg the Fukushima and Three Mile Island incidents both involved damage done by an inability
88.
by mlinsey 5y ago
"under the hood" can go arbitrarily deep; nobody thinks everyone needs a deep understanding of the quantum physics governing how a transistor works to build a web app. But in my experience, people saying "under the hood"
89.
by mlinsey 5y ago
Open platforms, where anyone can build anything without asking a central authority (with their own financial incentives) for permission, allow for much more innovation. As an easy example, take a look at the App Store and Apple's 30%
90.
by mlinsey 5y ago
A moderate spread of six story buildings might be optimal for energy usage of just the buildings, but total energy usage/environmental impact are also shaped by other human factors for where people want and need to live and work...eg w

More ›