josh-sematic

8 ms·

91.
Sand trafficking in Latin America(insightcrime.org)
108 points by josh-sematic 2y ago|56 comments
92.
Peter Thiel Company Generator(peterthielcompanygenerator.org)
2 points by josh-sematic 2y ago|0 comments
93.
by josh-sematic 2y ago
Yes. Hard to imagine a more hypocritical outcome of “free thinking” than “think like me or die”
94.
by josh-sematic 2y ago
I’m not an expert in such things, but this seems like a good use case for IPFS. Kinda similar to a torrent except that it is natively content-addressed (essentially the key to access is a hash of the data).
95.
by josh-sematic 2y ago
A common metric for how much actual content has changed is the Jaccard Index. Even for large numbers of datasets that are too large to fit in memory it can be approximated with various forms of MinHash algorithms. Some write up here: https
96.
by josh-sematic 2y ago
You might find this comparison chart helpful: https://www.airtrain.ai/blog/how-15-top-llms-perform-on-clas... Note: from October; also I work at Airtrain
97.
by josh-sematic 2y ago
TL;DR on the status: they've been working on it for a while, but it's not yet ready for playing around with.
98.
by josh-sematic 2y ago
I would say treating others good will, good faith, and sincere best efforts are my definition of respect. My guess is that if in your travels somebody informed you that not taking your shoes off at the door is a faux pas in their culture (p
99.
by josh-sematic 2y ago
Knowledge can inform external behavior. When you go to another country, you have to learn the norms of what will be considered disrespectful in that context. In America, for example, somebody waving their middle finger at you will be seen a
100.
by josh-sematic 2y ago
Are you suggesting that you think that disrespect is a better default behavior towards others than respect? For example, would you prefer that people respond to your comments disrespectfully until they have seen enough of them to believe yo
101.
by josh-sematic 2y ago
Treating people with respect can sometimes mean learning enough about them to understand a little about what life is like in their shoes. There are a lot of different kinds of people wearing a lot of shoes. Learning about them is a lifelong
102.
by josh-sematic 2y ago
Firm agree. To me the best way to build a mental model of some code is to be able to quickly answer the questions that arise in my mind as I read it. The order in which these questions arise differ for each reader, and indeed for each inten
103.
by josh-sematic 2y ago
If the trajectory of the past two years is any guide, things that can be done at great compute expense now will rapidly become possible for a fraction of the cost.
104.
by josh-sematic 2y ago
There’s clearly some value. People are paying for something. > AI start-ups generate money faster than past hyped tech companies https://www.ft.com/content/a9a192e3-bfbc-461e-a4f3-112e63d0b...
105.
by josh-sematic 2y ago
There are several allocation opportunities I could think of. You’re a local government considering some subsidies for rooftop solar initiatives. How much bang for your buck will you get? You’re a regional grid operator and have some estimat
106.
by josh-sematic 2y ago
Even assuming 100% solar rooftop coverage is the goal, given limited capacity of raw materials, labor, infrastructure would still necessitate prioritization of when to allocate those things to which places.
107.
GM is giving up on Cruise robotaxis, pivots to personal autonomous vehicles(techcrunch.com)
3 points by josh-sematic 2y ago|1 comments
108.
by josh-sematic 2y ago
Largely agree. Heavy process is a tool to help less skilled people have output more comparable to more skilled people. But it comes at a cost of making more skilled people less productive. If you can hire small numbers of highly skilled peo
109.
by josh-sematic 2y ago
Interesting! It seems to be that there would be a tradeoff between specialist subsystems (which allow you to excel at the specialized tasks, but which can't handle things outside the specialization well) and generalized subsystems (whi
110.
by josh-sematic 2y ago
As others have said I would say first-mover/brand advantage is the big one. Also their o1 model does seem to have some research behind it that hasn't been replicated by others. If you're curious about the latter claim, here&#
111.
by josh-sematic 2y ago
You’re not even close to hitting rock bottom on explanations of the layers below until you find yourself comparing proposed theories of quantum gravity :-)
112.
by josh-sematic 2y ago
One reason why I'm excited about o1 is that it seems like OpenAI have cracked the nut of effective RL during training time, which takes us out of the domain of just fitting to the curve of "what a human would have said next."
113.
by josh-sematic 2y ago
For what it's worth, here's the post I was referring to: https://www.airtrain.ai/blog/how-openai-o1-changes-the-llm-t... HN discussion here: https://news.ycombinator.com/item?id=41723384
114.
What Happened to 23andMe?(cnn.com)
4 points by josh-sematic 2y ago|1 comments
115.
by josh-sematic 2y ago
They are indeed similar and OpenAI did indeed use RL at training time in a way that has not been done before, as does this approach. Yes both also involve some additional inference-time generation, but the problem is that (at least as of no
116.
Lego robot controlled by artificial worm brain (2015)(ibtimes.co.uk)
1 points by josh-sematic 2y ago|0 comments
117.
Contentious California AI bill passes legislature, awaits governor's signature(reuters.com)
3 points by josh-sematic 2y ago|0 comments
118.
by josh-sematic 2y ago
If a question was created in 2013 and viewed by somebody in 2020, that moved the line higher in 2013, not 2020. A more useful view would be one where it contributed to 2020. In other words, the graph shown answers “how often were questions
119.
by josh-sematic 2y ago
Sufficiently large companies should have a healthy “internal portfolio” of projects as well. Some safe bets/easy wins, some “this probably won’t work but if it does it’s game changing” projects as well. If they want to continue to exis
120.
by josh-sematic 2y ago
Glad to hear Mozilla is working on adding this. I switched to a chrome-based browser for a while and the only thing I miss after going back to Firefox is tab groups.

More ›