l33tman

France managed to get in an exemption in the proposal for their police and security workers, who can keep their privacy. I think the wording is that this would only apply to apps available to "the public". So if "the public" can't download your app, you're safe...

I think it just shows it won't be possible to implement this in a useful way. Let's hope...

In the proposal, they write that the service providers have to figure out a way to make the false positives reported to the police "minimal". This is obviously a major burden to put on the service providers, in effect completely excluding all small outfits (and I'm sceptical the large companies want to deal with this either).

Germany have had some similar tech in place according to Der Spiegel, but the entire increase in positives was found out to be legal dickpics and flirty messages between teenagers etc. The only result was that the police now have a huge database of teenager's naked pictures and kids on the beach, which can hardly be a good way to minimize pedophile activity.

It will be a shitshow beyond comprehension if this eventually gets implemented.

Was travelling through Sassnitz / Rugen last year and the restaurants we visited didn't accept cards. We had to run around late trying to find an ATM the first evening. And that is a tourist region even..

Yes, some of the parties did. In Sweden for example, the government parties just decided to greenlight the EU proposal that is up for vote now, even though their EU parliament representatives (from the same parties) had said before the recent EU election that they were against it (they even celebrated publicly last autumn when the proposal was downvoted last time). It's a mess trying to keep track of the party politics to be fair and it's possible the EU parliament members are voting in one way while their "mother party" in the home land votes in another way. But it did feel like a rug pull here.

Tangentially, but I'm wondering how the calculations really go with the robo-taxis. Taxidrivers here are probably getting $5/h or so and it's a job you can get without any skills at all, you don't need to speak the language and you don't need any tests, you just need to be able to sit down, input stuff on a GPS and drive. So there's an infinity of people to employ if you're a cab company. Neither Waymo nor Tesla want to give the tech away for free. Is there a market (without corporate subsidies from Waymo etc)?

Yes I see the last one was supposed to flash twice, but at 2 positions. The explanation is the same though as far as I understood the article, that the brain processes all flashes as a single temporal chunk and that is what causes the temporal re-alignment. It kind of resembles the famous letter-jumbling where you can still read the text if you re-arrange the letters inside the words if you read fast enough.

Yeah it's supposed to be two flash positions only, but I missed that the last one is supposed to flash twice which it apparently doesn't in the gif so that was the confusion.

"The researchers exposed people to a distressing emotional experience for three days in a row: they had to listen to a recording of themselves singing—often out of tune—to karaoke, which aroused shame." Heh.. The researchers have been pretty creative :)

But yeah, I've experienced the same thing, you think you haven't slept a second during the night but you probably have

That is the whole point of the illusion, that there are two filled circles that flash and your brain perceives it as 3 flashes because they come so fast that they are processed as a temporal chunk before you become aware of it.

I still think it's a big difference between trusting existing OS'es and apps, which are under scrutiny by hundreds of security researchers and thousands of security nerds all the time, and willingly sending away all your data to a party who promises they will treat it well (I know it doesn't work like this in this case, but just for the sake of argument).

In essence, what you're doing is training an assistant to learn all of your details of your life and habits and the question is if that "assistant" is really secure forever. Taken to the extreme, the assistant becomes a sort of "backup" of yourself. But yeah it's an individual decision with the pro's and con's of this.

I think it boils down to that it doesn't matter what they promise, if you send a videocap of all you ever do on your computer to some company on the internet, you just have to take your chances. Would you put mics and cameras in all of your rooms in your home that send data to Apple (or someone else) to analyze "for your benefit" even if they say and promise they won't do anything bad with the feeds?

At least with gmail and chat clients etc. things are somewhat put in compartments, one of the services might screw up and do something with your emails but your Messenger or WhatsApp chats are not affected by that, or vice versa. But when you bake it into the OS (laptop or phone) you're IMHO taking a much bigger risk, no matter what the intentions are.

I don't think the above poster was really referring to who does it, but that it's creepy that you're having a conversation about your mom with your phone to begin with

Interesting summary from inside the article:

"Just a week ago, our field offices in Charlotte, Indianapolis, Jacksonville, Los Angeles, and Cleveland worked with the Defense Criminal Investigative Service and U.S. Secret Service—along with international partners from Denmark, France, Germany, and the Netherlands—to conduct a technical operation against four groups who offer malware as a service, in the first such operation ever conducted.

That operation, Endgame, defeated multiple malware variants, took down more than 100 servers, and dismantled the infrastructure for four key pieces of global malware, which had been responsible for hundreds of millions of dollars in damages and had even compromised the critical-care online system a hospital needed to keep patients alive. "

The ChatControl2.0 proposal was put forth by Ylva Johansson who is from the left in Sweden. She literally says "But think of the children" in every interview she made after this. There has been many accusations of corruption, for example the council was in contact with the AI companies that stand to gain from getting this in law.

https://www.svt.se/nyheter/ylva-johansson-ifragasatts-om-lob...

I don't like mass-surveillance or "dragnets" which are actually considered unconstitutional in the US (see https://en.wikipedia.org/wiki/Dragnet_(policing) ), where you make an extraordinary amount of people suspects in order to catch a single suspect.

In medicine there is a number called NNT (Number Needed to Treat) which is a really good comparision to this. It's how many you have to treat with a medicine or procedure in order to help one patient. If that number goes up too large it's a bad sign, like if you would have to feed 1 million people a pill every day to avoid 1 heart attack per year.

Here you have mass surveillance and an incredible breach of privacy of 500M citizens, in order to catch how many pedos? If they would even be caught by this at all to start with.

You put 0-ohms where you think you might at some point want to have something else there like a > 0 ohm resistor or an inductor (for EMI blocking for example), it's much easier to just reprogram the SMD robot than to make a new PCB.

But many chatgpt hallucinations are catastrophically wrong. For example when querying for things related to advanced scientific subjects in physics and medicine for starters from my own experience.

I really want to give this a pass but the worst thing is that you can't really see where it turns from correct to incorrect, like you would usually pick up with a human teacher or peer when you start to get a feeling they really don't know what they're talking about. ChatGPT will gladly keep hallucinating references and reasonings that sound superficially plausible until you look them up, because that's what it's trained to output..

If they could just find a way to cut it off when it starts to become too unsure, it would be a big improvement.

It isn't about copyright, it's about passing-off, it's described elsewhere in detail in these threads what it means. It's about intention and what the customer believes. If customers might believe its SJ, due to samas tweets, general likeness in voice, and the context (voice assistant), the public info about them trying to get SJ to do this - that's passing-off, even if it wasn't training on her voice per se. There are numerous law cases about this.

Fortunately for us customers, it seems Uber catalyzed a transition among most other taxi companies as well to improve this. A bunch of local taxi companies here have apps that have the same features as Uber, where you can pay in the app and see where the cab is on its way to you etc. There are "OEM" taxi callcenter cloud services that small taxi companies can rent to get the uber experience for their fleet without any IT knowledge, each car gets an iPad with an ODB connection and you're ready to go.

I've never trusted those predicted binding energies. If you have predicted a ligand/protein complex and have high confidence in it and want to study the binding energy I really think you should do a full MD simulation, you can pull the ligand-protein complex apart and measure the change in free energy explicitly.

Also, and this is an unfounded guess only, the problem of protein / ligand docking is quite a bit more complex than protein folding - there seems to be a finite set of overall folds used in nature, while docking a small ligand to a big protein with flexible sidechains and even flexible large-scale structures can have induced fits that are really important to know and estimate, and I'm just very sceptical that it's going to be possible to in a general fashion ever predict these accurately by the AI model with the limited training data.

Though you just need some hints, then you can run MD sims on them to see what happens for real.

My point one level up in the comments here, was not really that the system is broken, but more like asking how you can run these companies (google and that other part run by the deepmind founder, who I bet already has more money than he can ever spend) and still sleep well knowing you're the rich capitalist a-hole commercializing life-science work that your parent company has allocated maybe one part in a million of their R&D budget into creating.

It's not like Google is ever going to make billions on this anyway, the alphafold algorithms are not super advanced and you don't require the datasets of gpt4 to train them so others will hopefully catch up.. though I'm also pretty sure it requires GPU-hours beyond what a typical non-profit academia outfit has available unfortunately.. :/

That sucks a bit. I was just wondering why they are touting that 3rd party company in their own blog post, who commercialise research tools, as well. Maybe there are some corporate agreements with them that prevents them from opening the system...

Imagine the goodwill for humanity for releasing these pure research systems for free. I just have a hard time understanding how you can motivate to keep it closed. Let's hope it will be replicated by someone who doesn't have to hide behind the "responsible AI" curtain as it seems they are now.

Are they really thinking that someone who needs to predict 11 structures per day are more likely to be a nefarious evil protein guy than someone who predicts 10 structures a day? Was AlphaFold-2 (that was open-sourced) used by evil researchers?

Why DL generalizes well is still an open research problem AFAIK. I've read numerous papers that tries to argue one way or another why this works, and they are all interesting! One paper (that I found compelling, even though it didn't propose a thorough solution) showed that SGD successfully navigated around "bad" local minimas (with bad generalization) and ended up in a "good" local minima (that generalized well), and their interpretation was that due to the S in SGD, it will only find wide loss basins, and thus the conclusion was that solutions that generalize well seem to have wider basins (for some reason).

Another paper showed that networks trained on roughly the same dataset but initialized from different random initializations, had a symmetry relation in the loss landscape by a permutation of all the weights. You could always find a permutation that allowed you to then linearly interpolate between the two weight sets without climbing over a loss mountain. Also very interesting even if it wasn't directly related to generalization performance. It has potential applications in network merging I guess.

To say the least. It's a bit unfortunate that there is about 0 culture in the EU regarding moonshot projects compared to silicon valley. I've tried to get money a couple of times from government grants for (yet another..) foundational AI model, neuroscience inspired, but the grants instead seem to almost exclusively go to well developed industrial companies that now wants some free money to "leverage" ChatGPT in their existing internal processes.. and being still in the research phase, the more risk-averse VCs here are not touching stuff like this either.

So I guess what's left is doing these grand proclamations that you are going to "knock the crown off OpenAI" etc. Though, some sort of vision is good to have for sure :)

To put another perspective on this, lots of modern advancements in both ML/AI and especially computer graphics has come from ideas already from the 70-80s that were published, forgotten, and revived. Because underlying dependencies change, like the profile of the HW of the day. So just let the ideas flow, not every paper has to have an immediate payoff.

Other cars with self-closing trunks for example, immediately stop if they detect some resistance. I guess other carmakers care more about safety than tesla, at least for the cybertruck (thinking about the vegetable-chopping possible with their trunk).

I can see that if you have very little experience building and selling vehicles in colder climates you just think "oh so there can be ice, let's just allow it to break the ice with more force" instead of, like, not designing the door in that way in the first place.

I really recommend listening to the Darknet Diaries podcast (available on Spotify at least). Really high-quality interviews with both ex and current hackers, cybersecurity professionals etc.

Not related to the speed, but is there any terminal for Linux that works like the Mac OSX terminal, in that you can shut it down and restart and it will bring up all the tabs and their cmd histories and scrollbacks for each tab? They do that by setting different bash history files for each tab etc.

And I prefer GUI terminals for this use-case...

Isn't that exactly what they do if they have the stream-cutoff discussed in this thread then? At least implicitly..

It would be cool if there was like a threshold in training from scratch, where after this actually works for adding higher level knowledge. So you would start training it like usual to get it to absorb generic langauge skills and reasoning but save all the domain knowledge absorption for merging later

Regarding this "The suspiciously cheap $15 FPGA had equally suspiciously date and lot codes covered (white rectangles on the FPGA chip in the picture above). I have a development board of the same series chip with markings intact, so it definitely shouldn't look like this. It did end up working, but I wonder what the origin of these chips is."

I spoke with a GoPro operations guy once, and asked how you can get the chinese ripoffs at $40, that looked the same as their $400 "real" deal. He said that they literally get chips from scrapyards, like they get batches of memory chips that didn't pass the tests etc. They might work only in a very limited temperature range for example. Also many other components are picked and speced to work 6 months on avg not 6 years.. They know that most of those "gadget" products are bought and used for a weekend and then put in a cupboard anyway.

It would be fine for an R&D project like this but I would be very careful if I was going to make a commercial project (then you also have the politics side, probably fuelled by some paranoia - there are legislations coming up which prevents you from using chinese silicon components in products for certain markets)