Posts by Michael Wojcik 12271 publicly visible posts • joined 21 Dec 2007
Well, besides the reasons others have already noted – most significantly that his mouth wrote a check that his backers then had to cash – it's worth noting that Musk is addicted to tweeting. So obsessive, irrational behavior regarding Twitter on his part is not unexpected.
Hmm. I think I still have one or two Twitter accounts. Not long after Twitter became a thing I created a personal account so I could follow some friends and family members; I stopped using it when Twitter went to OAuth because that broke integration with the RSS reader I was using, and I wasn't inclined to switch to something else.
Then I created a work account some years later, on request from some Head Stooge of Social Media Wankage or some such office. I couldn't bear to follow the suggestion that I retweet marketing posts, though, so I never actually used that account for anything.
I never posted anything to either, though. I was briefly amused that I had more than thirty followers of my personal account despite it never having a single post. Silence is golden, I suppose.
So I guess I too will be there at the end, for an extremely small value of "there".
I don't think that's true for the machine learning methods used by Copilot.
Right. It's not a database (much less a relational one).
That doesn't mean it's necessarily impossible to create an explicable, interpretable, or self-interpreting Transformer model; but I can't offhand think of an approach which wouldn't require a lot more resources. The research I've seen into explicable/interpretable ML is mostly based on other architectures that are more amenable to extracting that sort of information.
I don't know that there's anything fishy with FTX itself. SBF found a good excuse for parting a large number of fools and their money, and then spread some of that money around. But much of it was based on the overall momentum of cryptocurrencies, and when that died down after the Terra collapse, it was just a matter of time before FTX proved to be overextended. It was a house of cards, and it collapsed.
What is fishy is what happened as FTX was circling the drain. The possible movement of resources over to Alameda, the "Bahamian residents" dodge...
FTX was just trading in snake oil for most of its history.
Molly White has details, as usual.
Only $4K/month? That seems low to me, based on the last time I talked to people about living in the Bay Area. But it's bad enough.
It's true this is very much a concern though. When my wife and I moved to Michigan, we were living in a lovely small city (only technically a city; it looked and felt like a traditional Midwestern town), in a historic Queen Anne that we bought for around $150K. A comparable house in SF currently lists around 10-20 times that (just checked some listings on Zillow). And we couldn't get our half-acre lot with mature trees in SF.
If I'd been living in SF, my salary could have been an order of magnitude higher and I would have been about breaking even.
what avantage is there over paying $250k/yr for two employees to work 8 hour days?
To be fair to the Muskinator, salary is not the sole cost of employing someone. The benefits an employer has to pay for a salaried employee in the US ain't cheap. And the Chief Twit is requiring everyone to return to the office, so rent, physical-plant, and other office-related costs will rise in proportion to the number of employees.
There's also a risk trade-off – two "moving parts" rather than one, to put it crudely – but on the other hand you also have a certain amount of redundancy with more employees, so I think that balances out under reasonable threat models.
It does seem like Musk could take this plan to its obvious conclusion and fire everyone except himself, and then pay himself 100% of Twitter's net profits. It'd probably work as well as his current strategy.
Agreed.
There was a time, quite a few years ago, when I spent a few months working 60-80 hours a week for my employer. But that was for a small startup where all the other (half-dozen) employees were family or friends, and they were also working very hard and often putting in long nights and weekends; and I was young and free of other responsibilities. It was temporary and we were doing what needed to be done to get a big deal and keep the company afloat, and to be honest, at that time and that point in my life, it was fun: the ergodic rush, the camaraderie, the sense of accomplishing something unusual and, in its own small way, notable.
If some jackass bought the company I worked for an announced I had to do that, I'd be gone as soon as I had another offer. Which, to be honest, if I worked for Twitter would have already happened.
It's an alignment-with-errors problem, and there's a ton of existing art and ongoing research, for example in genomics.
I've known CS grad students to build detectors for this sort of thing with good F0 and throughput, as class projects. If you're only concerned about checking specific Copilot outputs (e.g. those that show up for a set of queries you've defined) against a fairly small codebase (e.g. your personal projects), that ought to suffice. If you want to do bulk scanning you'll likely want to partition the datasets and do some preliminary classification to make the problem tractable.
Yes, I was going to post something similar. Copilot is in no way a pair-programming mechanism. The purpose of pair programming is vigilance, not copy-and-paste.
You could certainly train a model on common errors and have it flag those – though we already have on-the-fly static analysis without needing vast resources for a transformer architecture with a zillion parameters, so that would be Kind Of Stupid. But that would be something along the lines of mechanized pair programming. Not particularly good pair programming, since pair programming works best with agents that understand the intent of the code; but it would be closer to pair programming than Copilot is.
A number of cases along those lines have been described in legal blogs such as Underhill's Lowering the Bar.
It's a variation on the "trail of evidence leading to the suspect's door" theme. And that really happens, too; Underhill mentioned a guy who stole a lot of quarters from an automated carwash and left a trail of them as he tried to lug them all home.
The "tough on crime" movement in the Federal and state governments was equal-opportunity for legislators. Certainly many Democrats contributed – it was the Clintons who popularized the bullshit "superpredator" idea – but Republicans are by no means innocent.
Neither party has a consistently good record on race. Republicans initially were the abolitionist party, but many (including Lincoln) wanted to institute racial "purity" by deporting people of African ancestry back to Africa. And while Democrats were the main supporters of segregation and Jim Crow for decades, around the middle of the 20th century those roles began to shift with the Republican Party becoming more and more regressive while the Democratic Party officially supported civil rights. (If not particularly enthusiastically. Most of the civil-rights improvements in the US were a combination of popular support by the Silent Generation and decisions by SCOTUS prior to Nixon's appointment of Burger as CJ.)
And as I noted above, neither Clinton nor Obama were good on civil rights, broadly speaking; and Carter was hamstrung by rebellion in his own party, 70's malaise, a (largely failed) focus on environmental affairs, and foreign-policy distractions.
Maybe they thought your humorless reply to a joke was tiresome. Maybe they objected to Yet Another recitation of the prison-rape cliché.1 Maybe they similarly found your reference to AIDS gratuitous.
I didn't downvote, but I didn't think your comment added anything to the discussion, even though I believe the US carceral state is greatly excessive, abusive, discriminatory, and pretty highly placed even on our impressive list of national sins. Handwaving comments about sex crimes doesn't constitute a coherent critique of that system.
1That's not to suggest prison rape doesn't happen or isn't a horrible problem; just that trotting it out at every mention of incarceration does nothing to help, and in fact trivializes it through pointless repetition.
1. Because the printers still work, and there's no good reason to stop using them. My LaserJet 4M was purchased in 1991, which you may note came before 1995.
2. Because it does still work. Get the driver from Microsoft's official driver archive site, and you're good to go.
Honestly, while I've been using some form of vi as my programming (and HTML) editor of choice since the mid-80s, and use vim on both work and personal machines today, I wouldn't try to convince anyone to switch to it. I don't believe it's objectively better than other choices; it's just what I'm used to and matches my personal preferences. (I strongly prefer modal interfaces, and in find the anti-modal UI zealots obnoxious, since they refuse to believe everyone doesn't share their inclinations.)
Sure about that?
I recently got my 1991 HP LaserJet 4MP working with Windows 10. I had to download the driver from Microsoft's driver archives, which aren't easy to stumble across accidentally but can be located with suitable search terms. With that installed, it worked just fine in both PCL and PS mode, using a Centronics-to-USB cable I got off Amazon.
"Not Supported" with Windows often just means the driver won't be found and downloaded automatically. But it's likely in the archives, if there ever was one for modern (XP and later) Windows.
That said, I wouldn't discourage someone from using Liam's approach, which will likely be viable as long as the printer is still working, and may involve learning some new things in the process.
To be fair, Twitter is used by people in a lot of fields to keep up on news. It's very popular among IT security professionals, for example, some of whom will claim it's absolutely critical to their ability to stay informed.
I personally have never been convinced it's of real value in that domain, which is why I don't read Twitter. (I don't even bother clicking on Twitter links in articles and such, since Twitter requires Javascript, and allow-listing Twitter isn't worth it.) There are plenty of other security news sources that I think are adequately current – feeds and lists like Full Disclosure, email newsletters like SANS NewsBites and Bulletproof TLS, blogs like Schneier's and Ducklin's, and so on. But apparently that puts me in a minority in this field.
Yup. Lots of folks 'round these parts have solar ovens, since we get ~300 sunny days a year, and with the altitude and low humidity passive solar heating can be quite intense.
Granddaughter Major and I put one together out of cardboard boxes, aluminium foil, and an old window last summer. Needs some tweaking to be really effective, but it heats up enough to do some cookery.
Offhand, using 2.45 GHz seems like a bit of an odd choice, since lower frequencies would have less loss (not just from water vapor but other atmospheric components) and would allow rectennas with larger grid spacing, which in turn would use less material per unit area and allow more visible light through. But I guess that would mean a larger ground station, which in turn would wipe out the latter two savings? This is very much outside my areas of expertise, though.
Yeah, there aren't any frequencies in sunlight that aren't completely reflected by magical white shirts.
I'm not saying there isn't a problem, but this sort of argument is reduction to absurdity. "Microwaves" are typically defined as 300 MHz to 300 GHz, more or less. That's a pretty big chunk of spectrum, and there's some variation on the effects on living organisms over it. Sunlight is also not a single amplitude or frequency. The situation is more complicated than "25% MOAR BAD!!!1!".
Man, Wells Fargo get crap when they commit fraud themselves, and they get crap when they outsource it. Come on, people! What are they supposed to do?
That's been my experience too. Credit cards, like all the other retail-banking products and services I've used, have been better at credit unions and small banks than at big banks.
Of course a lot will depend on the individual institution. The credit union I used for the longest period of time was one of the biggest in the US, so they had a lot of resources and leverage. The small bank we're using now (after moving) has been a fixture of the community for a long time and a healthy balance sheet, so again it has decent resources.
Really. For retail banking in the US, if people can find a local small bank (good) or credit union (better) that meets their needs, that is such a better choice than a big bank.
Retail banking is a thin-margin business, and big banks are only in it for the opportunity to upsell. Small local banks actually have a financial reason to care about their individual (and small-business) depositors, and credit unions are owned by them.
I imagine [if] I were still in my teens/twenties... pretty sure I'd be logging on all the time
I'm not sure. Granddaughter Major is not quite to the teens yet, but has a phone and tablet and occasional use of the desktop machine. She does spend some time on FB Messenger Kids exchanging messages with friends, but probably most of her social device use is chatting while playing Minecraft. Her peers, from what I've seen, don't spend a ton of time on social media either. For them, reading and posting on soc-med forums is regarded as more of an adult activity. They'd rather be co-op gaming or similar, if they're online.
Lots of BBSes had Internet or email gateways too. UUCP was still useful for quite a few things that weren't on the Internet. WAIS was useful if you had access, as many university libraries did; and a lot of library catalogs had Telnet-based online access. I did quite a bit of online research at Northeastern University in 1990-1991.
1991 was also the year of CIX and ANS CO+RE, and then in 1992 the NSFNET AUP was superseded to allow commercial traffic on the NSFNET backbone, so we were right on the cusp of the commercial opening of the Internet (for good or ill).
Indeed. The Internet in 1991 was still mostly for research, socializing, and software distribution, and it worked well. Usenet (then still a mix of NNTP and UUCP) wasn't perfect, but it was noticeably superior to pretty much all of its successors. MIME was still 5 years away, so email was still good. The web was barely getting started (CERN httpd was released at the very end of 1990), but there was plenty of useful and interesting content; files were located with archie and downloaded using FTP.
Security was still lousy—this was only three years after the Morris worm—but attacks were still pretty rare. Eternal September was a couple of years away (1993) and widespread Usenet spam ("Global Alert for All" and "Green Card Lottery") a year further than that.
I'm not saying I want to go back to the Internet of 1991, mind. There's a ton of stuff I do like available on the web. I even play a couple of games on my phone once in a while, and I appreciate the ability to read a book on it even more. (I used to always carry a paperback when I left the house, in case I had to wait somewhere; now my phone serves that role. It's convenient.) But I wouldn't be terribly sorry if 1991-style Internet was all we had now, or even if I did have to go back to that.
Urgh. The money would be nice (should I ever find myself in that position), but I can't imagine spending half a year sitting around with no job and not trying to get one. I have plenty to do around here—and plenty more I could do, like excavate a root cellar and turn the excavated material into adobe bricks to build another outbuilding—so I wouldn't be playing games anyway. But that'd still drive me batty.
But to each his own.
RTFA.
AI Hardware Summit in September, where the company showed off analog chips that could run the YOLOv5 object detection algorithm on high-resolution video at 60 frames per second while only consuming 3.5 watts
So they have, in fact, shown something.
Power consumption is gradually becoming a bigger and bigger concern for large-model ML. But hardware choices move a lot slower than software in the field, and the big consumers mostly already have a lot of still-usable hardware in data centers, so they wouldn't be in the market for something like this yet. And rewriting ML software to work with a different architecture could be expensive.
Really low-power ML solutions are more tempting for embedded and "edge" (ugh) applications, so there is a potential market for something like this in applications such as self-driving vehicles. (Whether that's desirable at all is another question.) But again the cost of switching to a new architecture is a big barrier.
Wealthy? Does that mean she gets to keep the proceeds of her fraud?
From the point of view of a legal team on contingency, she's wealthy until she runs out of appeals. Which is what's just happened, it appears.
Honestly I don't have a problem with that, either. She's entitled to a defense. I'm not questioning the verdict, and I suspect if I'd been on the jury I would have voted to convict as well.
But I also agree that we need better funding for public defense, to reduce inequity in the system. And we need to cut down on prosecutorial abuse of the plea-bargain system, and on the "chickenshit prosecutor" phenomenon that discourages prosecution of wealthy and powerful offenders.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
