Very cool! I wish this existed a couple years ago. I was hacking with LSP and a Java language server for a code migration prototype, and I could not even get past the initialization stage. The server (must have been the Eclipse one) didn't seem to respond to the initialization params as expected, and I could not find the right documentation or code snippets to figure it out. I eventually implemented the prototype using OpenRewrite. With this, I'm tempted to revive that idea again.
This was precisely the situation I was in! Luckily, the Eclipse JDT.LS contributors are super helpful, and provided me with a lot of their time answering all my questions, which I have now tried to document in as much detail as possible in the Eclipse part of multilspy. My sincere hope is that multilspy can serve as the repository for all other language servers.
I hope you revive "that idea" and I would be glad to help in any way possible w.r.t. multilspy to help you through!
"... a new paper shows GPT-4 simulates people well enough to replicate social science experiments with high accuracy.
Note this is done by having the AI prompted to respond to survey questions as a person given random demographic characteristics & surveying thousands of "AI people," and works for studies published after the knowledge cut-off of the AI models."
"Applying Asch's conformity experiment to LLMs: they tend to conform with the majority opinion, especially when they are "uncertain." Having a devil's advocate mitigates this effect, just as it does with people."
'But Judge Lewis Kaplan called Trump’s semantic argument “entirely unpersuasive.” He clarified that the jury found that the former president did indeed “rape” Carroll based on the common definition of the word.'
Here’s an alternative take. Note, I’m absolutely not a Bezos fan. Maybe he is just chickening out against Trump and/or fighting against the proposed billionaire tax (which IMO will never happen.)
But consider that 1) even with all the damage Trump could do, Bezos will still be richer than god, 2) Bezos did not instruct the Washington Post to endorse Trump, and 3) he doesn’t seem to have asked them to keep things quiet either.
So of course this story breaks and of course there is all this media hullabaloo with the upshot being everyone now:
* knows that the WaPo was about to endorse Harris.
* is reminded that Trump has made official decisions and improperly pressured government matters based on personal feelings.
* is aware that even the 2nd richest man in the world fears the personal ire of a presidential candidate in a democracy, ostensibly with a solid rule of law.
I hate that this comes across as “he’s the billionaire we deserve, but not the one we need right now, and oh, BTW he's also playing 4D chess," but all this seems very expected. So maybe another way to look at this is: Bezos appears to submit to Trump, which in itself serves as a very publicly warning to the world about what will happen under Trump, and indirectly endorses Harris anyway.
I don't think he has to do 4D chess to signal support against Trump. He could always just use his giant megaphone to decry Trump as a wannabe dictator that improperly used his own power when he was President to punish his political enemies - something Trump does not shy away from publicly admitting.
It’s one thing to be another voice in the crowd of people all saying the same thing (even a very deep pocketed voice) and another entirely to be the object example everybody discusses by letting it be believed that with all your
billions, you’re scared too. People are more affected by things they decide for themselves than things you tell them, hence show don’t tell.
Were that to be the case, I wouldn’t consider it 4D chess. It seems like a straightforward strategy to me.
> (though I still don't appreciate how this would practically be done to the existing hardware, and they apparently only simulated this with software-defined radio).
It is my understanding that most modern baseband chips can effectively be considered "software defined radios", as most of the modulation/demodulation is performed by the firmware. While the researchers appear to have used a USRP (a dedicated SDR platform), it is conceivable their scheme could be accommodated in the firmware.
As far as I know, transmitting and receiving Wi-Fi traffic will never be possible using GNUradio, because you cannot meet the maximum 16 microsecond latency for sending an acknowledgment after you decoded a packet successfully.
I'm not sure what you are asking it should be comparably priced to, but USRPs are on the higher end of the cost spectrum. Caveat: my experience here is extremely limited, but at one point I too was looking into affordable GNURadio-compatible SDR hardware that could transmit and receive (as opposed to the RTL-SDRs that can only receive) and I came across options like HackRF and LimeSDR.
However, knowledgeable people also pointed out that these cheaper options make tradeoffs in the RF hardware that make it harder to get reliable performance for non-trivial uses. Their opinion was that the time saved in working around those limitations was well worth the extra cost of a USRP.
The BladeRF and ADALM-PLUTO are cheaper alternatives to the entry level (B2xx) USRPs. They use the same Analog Devices MIMO chip as the USRP, so are similar in capability.
> in an important sense, society itself is a form of advanced intelligence
This made me think of Charles Stross' observation that Corporations (and bureaucracies and basically any rule-based organizations) are a form of artificial intelligence.
Very interesting, I'm working on exactly the same problem from a couple different angles, but I'm not having much luck. I have negligible background in AI/ML or computer vision however, so I'm most certainly Holding it Wrong (TM). My general approach has been trying to generate embeddings using smaller models like MobileNet and ResNet (not trained or finetuned or anything) and using similarity metrics like Cosine distance, but there's too many false positives. If you can disclose it, would you be willing to expand on what has worked for you?
reply