If GOTW, actually means GWOT and Global War on Terror, please have courtesy of reading your posts beforehand, and perhaps also not assuming a US audience.
Dwarkesh, is one of the better tech podcasters.
Scott is well known for blogs Slate Star Codex and Astral Codex Ten.
Daniel resigned from OpenAI in 2024,and is known for prescient AI boom predictions made in 2021.
https://www.alignmentforum.org/posts/6Xgy6CAf2jqHhynHL/what-...
Today they jointly published their 2027 predictions
https://ai-2027.com/
Having lived in UK, France and Spain, I know that there aren't daily variations in consumer electricity prices, as there are in places like California. I don't have data if offices are priced differently.
When you say 'defining languages', do you mean as DSLs or choosing appropriate existing languages? If DSLs, which language would you recommend for creating DSLs. Could you give a successful example?
I think the move Rowson's referring to is move 9. Since in the essay he says
"Yermolinsky offered a pawn as bait, and I very nearly didn’t take it because doing so would allow him to play a series of forcing moves"
So the only pawn that was offered and that Rowson took was on move 10 in reply.
The only reason I checked was to see how far a grandmaster was looking ahead, in this case 14 moves, which i find impressive.
"...I discovered a surprising detail right at the end of the line, in which my knight could retreat back to its original square, "
Like you, I wouldn't be impressed if he only saw from move 18 to move 24.
Pure speculation, isn't the most interesting time period in the signal the transition from signal to no-signal (and back again)? How high is the resolution of the sensors? Is it possible that some frequencies are blocked fractionally earlier than others? At a fine enough temporal resolution, there might be more clues regarding the source.
Yes, I know. But the tone of the whole article, is as if, they've found deep flaws across many VMs. They call something 'warmup" which I think has little or nothing to do with the JIT, but is unaccounted variations in the whole running system.
The final graph shows a binary trees program in C, with a 6% variation between "in process executions", and no steady state, it seems logical that most VMs will show the same or worse variation.
The "warmed-up steady state" does exist, but not if they define it so narrowly. All of their iterations and timings are running at x30 to x100 interpreted speed, the only 'cold' interpreted code is in a few microseconds of the first loops of an execution.
It seems that the authors are not measuring what they think they are, or have explained it poorly. Most transitions from interpreter to JIT show speedups of x10 to x100, eg luajit or V8. How is it possible that the variation of V8 (as an example), according to their numbers is showing improvements of only a few percent, when it should be orders of magnitude faster after transition? My conclusion, they are measuring variations after warmup.
All of the warmup, and transitions from interpreter, to JIT, to optimised JIT , happen inside the first few micro or milliseconds of EVERY one of their thousands of process iteration. Their measurements are ALL of the system variation of the VM after warm up has taken place. The VM is optimizing within the first 1-1000 inner loops occuring at the start of EACH process iteration. For most working programmers, a variation of a few percent on a running system AFTER warm-up in "steady-state peak performance", and before any I/O takes place (because language benchmarks avoid I/O), would not be an issue. If it is an issue, then the article perhaps demonstrates that a compiled language would offer less variation.
The benchmarks listed range from a shortest of around 0.4s for fannkuch/hotspot/linux, up to 1.8s for n-body, pypy, linux. This 'long-running' benchmark code (of .4 to 1.8s ), by definition, has to include multiple inner loops/hot code, which is quickly optimized, otherwise benchmark code would have to be millions of lines long, in order to have a sufficient runtime length. Tests need to run for at least tenths of a second, for cross language comparisons, since JITted languages take some iterations to warm-up.
I've googled generally and verbatim with various quotes from above and your name, and cannot find the essay. It sounds interesting. Can you post a link please.
The same real world react app went from 40MB to 20MB.
They don't ship minified javascript to final app, but bytecode, which they say is 1/3 smaller (both minified). They have moved the parse and compile stages to ahead of time, making the VM smaller, and giving the bytecode benefit. The bytecode benefit also means a quicker time to live/first reponse from 4s to 2s in their example.
Link to the talk, from FB employee announcement made at React Conference today, 15 minutes length. Linked 2 mins in, because the audio is poor to start.
Why are you assuming other renewables is mostly waste/biomass? In the notes below the table, wind is cleary mentioned.
My guesstimate would be that wind was a large proportion. I confirmed by checking wikipedia.
EIA Total for year was listed as 4,177,810 in 1000s of MWhr
Other Renewables 354,445 in 1000s of MWhr
2018 Wind power generated 274,962 in 1000s of MWhr. So makes up over 75% of other renewables. Chart shows wind has grown from 94,000 in 2009. Which combining with the EIA data, showing 143,388 in 2009, means most of the growth in non-solar renewables is from wind.
At 12km/h it requires 5.59KW of sun energy to fall on the car (in total, not per sq meter as in original comment).
Actual figure given on website is 5m2 of panels on car. Assume 1Kw per m2 of sun power (near equator, no cloud, midday) @ 22% conversion efficiency gives 1.1kw electrical output from panels.
So one needs 5.59KW sun power input for 1.1kw output.
We've both read the article. On rereading the article, its not clear if the stolen ID was present at all the crimes. My assumption was that it was not.
I make that assumption, because the person accussed of multiple crimes is suing Apple based on their use of facial recognition. He wouldn't do that if the ID was used in each crime.
Another article confirms my assumption is correct.
"A detective with the New York Police Department allegedly told Mr Bah that the thief probably used Mr Bah's driving licence as identification during one of the robberies. The detective reportedly said that this may have caused Mr Bah to be charged with thefts committed at Apple Stores in New York, Delaware, New Jersey and Massachusetts, according to court papers."
Apple took the stolen ID used by the criminal. Then Apple incorrectly linked the details of that ID (with no photo), with images of the thief.
Following further thefts associated with that ID, the innocent man was accused of all the thefts based on the ID and an INCORRECT photo match to the stolen ID which Apple had created in its facial recognition system, and used as the basis to inform the police of the suspect.
I tried my best not to be snarky. I thought the article was a normal average article.
Edit: I'm not sure the article says facial recognition was responsible, I think that's what the courts will decide. But to me, its clear that facial recognition was involved, based on the article's title and content