thaw13579

6 ms·

31.
by thaw13579 1y ago
I think a better analogy is large corp built & owned vs small artisan businesses & single family homes. Why not have both?
32.
by thaw13579 1y ago
This use of the term "fiction" is mostly popularized by Yuval Noah Harari, and while correct in some sense, I find it tends to favor its meanings of falseness and unreality. Another more widely used term is "social construct
33.
by thaw13579 1y ago
Life expectancy was shorter a century ago because infant mortality, disease and injury pull the average down. We've done an amazing job of a society pulling up that lower end, but lifespan associated with normal aging is actually fair
34.
by thaw13579 1y ago
Lately people with high a level of wealth and power like Ellison view themselves as a privileged class and have no pretense that their machinations should apply equally to all. The erosion of that norm, equality of all under the law, is th
35.
by thaw13579 1y ago
Hopefully they're solving for better social connection, not solving for chronic loneliness as so many other for-profit ventures have...
36.
by thaw13579 1y ago
Autism isn't as cut and dry as this. Falling out of planes and diabetic comas are truly black-and-white outcomes, but ASD assessment involves behavioral testing of social and verbal ability, which has high variability between and with
37.
by thaw13579 1y ago
I reviewed the linked studies of folinic acid treatment in ASD, and they uniformly say that larger studies are needed before considering this as a widely available treatment. The main issues that need to be worked out are: * who this appli
38.
by thaw13579 1y ago
This has been my observation as well. To add, I'm seeing leadership and stakeholders use their chats with LLMs to justify claims like "what I'm asking for is incredibly simple according to ChatGPT, and it should be done by e
39.
by thaw13579 2y ago
My sense is that this was written with an in-person team that in mind. For remote teams, daily standups are also a way to “sit together” like the author suggests.
40.
by thaw13579 2y ago
The first turn in an S-curve can easily look like an exponential. ASI has physical limitations, so I don’t see why it wouldn’t take an S-curve as well, although at a much different rate than human intelligence.
41.
by thaw13579 2y ago
I think the parent is talking about banning them specifically in pedestrians and cycling lanes. We currently ban motorcycles and cars from using them, presumably because it’s dangerous to pedestrians.
42.
by thaw13579 2y ago
Performance and cost are trade-offs though. You could just as well say that LLMs are dead in the water, except in terms of performance. It does seem likely we’ll soon have cheap enough LLM inference to displace traditional NLP entirely, al
43.
by thaw13579 2y ago
For many people in cities, it’s hard to secure parking with reliable access to an electrical outlet.
44.
by thaw13579 2y ago
My highly opinionated take from working in a field with replication issues -- the people publishing unreproducible results simply want to either establish or reinforce high social standing. They are highly intelligent and skilled in the sen
45.
by thaw13579 3y ago
It depends, the compensation is highly skewed. The PI in question is likely compensated over $1M/year, while other investigators involved (assistant and associate professors) have compensation comparable to entry-level programmers. T
46.
by thaw13579 3y ago
While that seems reasonable, it assumes the amount of impact per paper has remained constant, while is practice it has greatly declined. These days scientists will literally try to identify the set of "least publishable units" in
47.
by thaw13579 3y ago
Counterpoint, in the context of education, the goal of the assignment isn’t to produce the end product of a 3000 word essay, but to have you go through the associated process of formulating, organizing, and presenting your thoughts. LLMs g
48.
by thaw13579 3y ago
Beyond this, some reviewers use their position to take advantage of the process to suppress, slow, or preempt work that overlaps or disagrees with their own. The journal editors are supposed to police this kind of thing, but they are simi
49.
by thaw13579 3y ago
One difference is that humans are actively involved in data collection, so when there is a gap in their knowledge, they don’t just wait for the information to show up, they ask a question, etc.
50.
by thaw13579 3y ago
Much of that time also includes physical interaction with the world, which makes it far more valuable because it can improve performance in a focused way.
51.
by thaw13579 3y ago
FWIW, I took “GPUs are deterministic” to mean they are deterministic in all possible intended use cases. This is not strictly true, since the whole point of using them is massive parallelism, which brings along non-determinism, for reasons
52.
by thaw13579 3y ago
I think whether it’s a bug or not depends on the software requirements and expectations. If the code has some expected bounds on runtime, switching the GPU code to sequential processing (for the sake of exact reproducibility) would break t
53.
by thaw13579 3y ago
Thanks for putting this together and sharing
54.
by thaw13579 3y ago
Papers are often rejected due to lack of novelty compared to previous work, especially if they contradict it. How many paper were similarly rejected because this fraudulent work was published?
55.
by thaw13579 3y ago
Seems meaningless without more information about experimental design. There are many ways to measure happiness and well-being, and many potential confounds by demographics, as jobs are not assigned randomly. For example, age and social ne
56.
by thaw13579 3y ago
They’re useful for tasks that are easy to verify but hard to generate, and code is one meaningful example. It seems trite to say after so many others, but I built a web app in one day by pair coding with GPT-4, which I am sure would have t
57.
by thaw13579 3y ago
What’s innovative about the LLM chat interface is that it requires no technical expertise to use. Aside from logging in, it’s not much different than sending a text message or email, so most people could experiment with it if they were at
58.
by thaw13579 3y ago
Yes, the goal post seems to move quickly with LLMs. Search engines provide relevant documents, not truthful ones, and I don’t know a single person who isn’t occasionally wrong, often confidently. I suppose it’s good to set a high bar with
59.
by thaw13579 3y ago
For me, it’s already a jetliner in the context of coding assist. It’s correct more often than top hits provided by a top search engine (and any coworker), and it is a very enjoyable user experience (no ads and SEO garbage to filter out).
60.
by thaw13579 3y ago
Today’s LLMs like ChatGPT also might seem like mainframe computers compared to the personalized and locally-run models that are likely in the future.

More ›