The best part of shopping at IKEA (which I've only done a few times) is seeing who looked at the map to figure out which unlabeled doors to go through to shortcut the maze.
Posts by Michael Wojcik
12317 publicly visible posts • joined 21 Dec 2007
Page:
- ← Prev
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 20
- 21
- 22
- 23
- 24
- 25
- 26
- 27
- 28
- 29
- 30
- 31
- 32
- 33
- 34
- 35
- 36
- 37
- 38
- 39
- 40
- 41
- 42
- 43
- 44
- 45
- 46
- 47
- 48
- 49
- 50
- 51
- 52
- 53
- 54
- 55
- 56
- 57
- 58
- 59
- 60
- 61
- 62
- 63
- 64
- 65
- 66
- 67
- 68
- 69
- 70
- 71
- 72
- 73
- 74
- 75
- 76
- 77
- 78
- 79
- 80
- 81
- 82
- 83
- 84
- 85
- 86
- 87
- 88
- 89
- 90
- 91
- 92
- 93
- 94
- 95
- 96
- 97
- 98
- 99
- 100
- 101
- 102
- 103
- 104
- 105
- 106
- 107
- 108
- 109
- 110
- 111
- 112
- 113
- 114
- 115
- 116
- 117
- 118
- 119
- 120
- 121
- 122
- 123
- 124
- 125
- 126
- 127
- 128
- 129
- 130
- 131
- 132
- 133
- 134
- 135
- 136
- 137
- 138
- 139
- 140
- 141
- 142
- 143
- 144
- 145
- 146
- 147
- 148
- 149
- 150
- 151
- 152
- 153
- 154
- 155
- 156
- 157
- 158
- 159
- 160
- 161
- 162
- 163
- 164
- 165
- 166
- 167
- 168
- 169
- 170
- 171
- 172
- 173
- 174
- 175
- 176
- 177
- 178
- 179
- 180
- 181
- 182
- 183
- 184
- 185
- 186
- 187
- 188
- 189
- 190
- 191
- 192
- 193
- 194
- 195
- 196
- 197
- 198
- 199
- 200
- 201
- 202
- 203
- 204
- 205
- 206
- 207
- 208
- 209
- 210
- 211
- 212
- 213
- 214
- 215
- 216
- 217
- 218
- 219
- 220
- 221
- 222
- 223
- 224
- 225
- 226
- 227
- 228
- 229
- 230
- 231
- 232
- 233
- 234
- 235
- 236
- 237
- 238
- 239
- 240
- 241
- 242
- 243
- 244
- 245
- 246
- 247
- Next →
Cardboard drones running open source flight software take off in Ukraine and beyond
Twitter scores legal hat trick with three cases filed against it in one day
Canada sticks a privacy probe into OpenAI's ChatGPT
Re: A Logic Named Joe
Shaun Nichols, formerly of this parish, wrote a retrospective on "A Logic Named Joe" for the Reg back in 2016. It includes a link to a Google Books page where I think you can read the whole thing.
ChatGPT becomes ChatRepair to automate bug fixing for less
Re: That's just great, Ollie
To be fair, frequently whoever wrote the original code didn't understand the problem either.
Testing assistance can be useful. That's what fuzzers do, after all. If an LLM-based system can generate valid unit or functional tests, that's one of the few areas where I'd be happy to have it. But suggesting fixes? Sure, let's avoid making developers learn anything, or think about why the code is wrong in the first place and perhaps find occasions for remedying anti-patterns, refactoring duplicated code, or providing better defensive mechanisms.
The vast majority of papers and comments I've seen in favor of using LLMs in software development (from people across the AI enthusiast-skeptic spectrum) boil down to "I'm a lousy developer and I don't want to think about what I'm doing". Of course they typically use the euphemism "more productive", which is the scripture most often quoted by the devil in this field.
As defense tech goes commercial, does national security miss out?
Hey Siri, use this ultrasound attack to disarm a smart-home system
School principal resigns after writing $100,000 check to Elon Musk impersonator
Re: Usenet
Dunno about Usenet four decades back — I started on Usenet a mere 32 years ago. But I used my real name then, and I used it on smaller social-media predecessors such as BBSes and university message-posting systems in the 1980s.
Many of the big-name Usenet personalities in the 1990s used what appeared to be their real names, or at least wouldn't go to any length to hide them. Everyone knew Kibo was James Parry and snopes was David Mikkelson. And in many of the Usenet groups I frequented there were people of local fame who used their real names. rec.arts.sf.written is an example; folks like Charles Stross and Pamela Dean Dyer-Bennett and Jo Walton appeared as themselves, if memory serves.
Exactly. It's the people who confidently declare they won't fall for scams who worry me. Human cognition is rife with pitfalls.
Pessimism and paranoia help, but they have opportunity costs. I could be much wealthier if my appetite for risk were higher and ethical reservations lower; I have acquaintances who have backgrounds similar to mine and are not so burdened who are indeed much wealthier than I am. I could also be much poorer. And I don't make the mistake of believing I can recognize all scams.
So you want to integrate OpenAI's bot. Here's how that worked for software security scanner Socket
Re: Obfuscated code
The problem is that you can use a GAN approach: one LLM acting as a detector, the other as a generator. You just invert the objective function so the generator refines its output until the detector fails to detect the malware. It needn't be obfuscated in the sense you're using the term, just too complex for the detector to ferret out.
Version 100 of the MIT Lisp Machine software recovered
Re: RISC Machines
Yes. And it's a bit odd to say "RISC is enjoying a renaissance" without noting that the vast majority of personal computers — smartphones, of which there are ~10 times as many as PCs of various stripes — are all using RISC CPUs.
And, of course, x86 CPUs have RISC cores, but Liam more or less alluded to that.
Re: Good or just entertaining
Your argument assumes LISP was designed ab initio under a deliberate "the right thing" approach. Obviously that's not true. It wasn't even originally intended to be implemented, and the REPL came along later more or less on a whim. But that doesn't prove that subsequent LISP development wasn't largely guided by a TRT philosophy.
The histories of LISP variants on the one hand, and of UNIX and C on the other, are of course far more complex than Gabriel's article even touches on. His piece is at best an interpretation of certain prominent aspects of those histories — and at worst an utter fiction; I'm not interested in trying to support any position along that spectrum. But in any case, the existence of anecdotal nits like the etymology of car and cdr (or other LISP awkwardness, like some of the special forms or the over-reliance on arbitrary punctuation or some of the infelicities Gabriel himself mentions) is a feeble counter to his argument.
Re: Good or just entertaining
No, no. Gabriel says in "Worse is Better" that he's deliberately setting WiB up as a strawman but then arguing that in fact it's more successful.
I have intentionally caricatured the worse-is-better philosophy to convince you that it is obviously a bad philosophy and that the New Jersey approach is a bad approach.However, I believe that worse-is-better, even in its strawman form, has better survival characteristics than the-right-thing, and that the New Jersey approach when used for software is a better approach than the MIT approach.
In effect: "Here's the approach I worked under, and here's the competition. I like mine better, but the competition is more successful and arguably better for developing software". Then he argues that latter point by noting how easily C and UNIX spread (the "virus" argument), and how that in turn attracts resources to improve them.
Ex-politico turned Meta hype man brands Metaverse 'new heart of computing'
EU mandated messaging platform love-in is easier said than done: Cambridge boffins
And what if you just want to have one ID on one system instead of buying into every service every would-be tech bro sets up
Maybe people don't want a single ID on all systems. I don't.
Maybe this "problem" isn't a concern for many users. It certainly isn't one for me.
Maybe the EU should keep its fingers out of domains it doesn't understand.
Re: What a load of bollox
The target outcome is that everyone only ever needs one messaging app
The "target outcome" is that no one has end-to-end encryption.
Honestly, I don't understand why some people find this so difficult to comprehend. If you force OTT messaging apps to interoperate, you lose privacy and security.
Judge grants subpoena to ID Twitter source code leaker
Re: Say 'cheese'
Better would be to run it through various filters and leak it piecemeal. The filters should be specific to the languages used and know how to vary things that don't matter, such as whitespace and other formatting aspects, and capitalization in comments. Depending on your purpose, it might be reasonable to do some obfuscation of identifier names.
Re: Say 'cheese'
Yes. There are any number of proposals for steganographically watermarking plain-text documents, including by spacing and other variations. For example, with source code you could inject typos into comments. The principles involved were being discussed at least as far back as the early 1990s.
64 points where you can make a single change gives you 2**64 identifiers. Personally, I'd go with more – perhaps 512 points of change, so I can use a group or erasure code to recover the identifier even if some of the changes are removed or obscured.
It wouldn't be hard to hack this mechanism into a git or other SCCS server, based on the identity of the person checking out the code. Of course, you still have to prove operation; the identified user could claim their account had been hijacked, for example. That quickly gets you into technical weeds where a judge or jury are unlikely to have expertise, so it's a battle of expert witnesses.
Re: Copilot
Gosh, I'd love to see a protracted legal battle between Twitter and Microsoft over that. With a little luck Free Speech Twitler would rage-quit (as he generally does when faced with legal complications) and Twitter would be finished off, while Copilot would get a reputation as a legal minefield. Everyone wins.
Nostalgic for VB? BASIC is anything but dead
Re: I think I can safely say ...
I've worked with a great number of programming languages, in all the various families (procedural, OO, functional, dataflow, etc) from various assembly languages up through 4GLs and pick-and-place graph languages. I don't think any of them inherently do a good job of teaching how to program well, much less how to be a good software developer (much of which is not programming), regardless of how frugal or profligate they are with resources.
Languages simply do not have that much influence. You can write structured, robust, maintainable code in any language, even something really awful like the Praat scripting language or Perl. You can write terrible code in any language, even one with a lot of expressiveness and syntactic sugar and guardrails and hand-holding. Yes, it's easier to learn data structures in Pascal than it is in traditional BASIC. Yes, it's easier to learn functional programming in LISP than in COBOL (in fact it's a real pain to do it in standard COBOL, but that doesn't mean it can't be done). But no language forces people to learn good techniques and habits. I don't believe any language forces them to commit to poor ones, either.
US cyber spymaster calls TikTok China's 'Trojan horse'
Re: Authoritarians...
Education, yes, but you have to climb Maslow's ladder. Most people have to have their basic needs addressed before they have much capacity for or interest in education. For a "rich nation" the US has far too much food insecurity, housing insecurity, and so on. The sky-high incarceration rate and dangerist War On Parenting are also problems. Dumping domestic problems on a completely unsuited police force is a grievous error.
Then when you get to the school system, you have chronic underfunding and anti-intellectual lunatics like DeSantis trying to burn everything down.
TikTok's pretty low on the list of actual problems, and if we're raising a generation that feels a little shittier about America, well, we deserve most of the blame.
Sure, it's a protectionist stance, there's a strong undercurrent of Orientalist1 racism, and TikTok is being punished for success. All true.
That doesn't mean TikTok isn't moderately to severely awful, or that it's not a problem that a great many young people in particular are obsessed with it, or that it can't be used to project "soft power" in ways that are favorable to China and unfavorable to the US. Someone can be right for the wrong reasons, and not all of the reasons here are wrong.
Personally, I'm a strong proponent of freedom of expression and still unwilling to agree with a TikTok ban. But I recognize that it's pretty toxic. (I did, however, enjoy Ryan George's YouTube piece about TikTok, so, y'know, feel free to call me hypocritical.)
1In the sense of Edward Said's most famous work, though the "Orient" he was primarily talking about was not the Far East.
For whom the bell polls: Twitter voting is for Blue users only now
Re: I voted years ago..
Agreed. I've seen quite a few people working in some of my areas, such as software security, claim Twitter is necessary or invaluable or terribly important to their work. I don't believe it. I've never seen anything, found out it had appeared on Twitter earlier, and regretted not seeing it then. It Just Doesn't Matter.
And the Reg only picks up some of the biggest stories. I've been reading Twitter is Going Great (inspired by Molly White's terrific Web3 Is Going Great, of course), and there are Just. So. Many. incidental anecdotes of Musk's hypocrisy, lies, and mistakes. Personally, as a Twitter refusenik, I'm finding it all quite entertaining.
Re: I can see the value...
After all, it might just make the bottom feeding trolls think twice about harassing people if they can actually be more readily held to account.
Certainly worked for SomethingAwful! Oh, wait.
Which is not to say that SomethingAwful wasn't a remarkable cultural phenomenon in its own right, and in places far more entertaining than anything I've ever seen that originally appeared on Twitter (according to citations, since I stopped reading anything on Twitter after just a month or so). Or that SA's pay-to-play system wasn't successful under its own terms. Or that there were no consequences – 4chan, for example, was created by and for people who were booted from SA. But there was plenty of bad behavior on SA; the goons more or less invented doxxing, for example. And that's to say nothing of the complexities and tragedies of Lowtax's own story.
Google again accused of willfully destroying evidence in Android antitrust battle
Yes, the chat exchange quoted in the article was marvelously stupid. Particularly since they could have just added a lawyer to the chat and let it fall under privileged communication. Privilege isn't bulletproof but it offers significant protection against discovery.
Personally, if I were running Google ... well, they wouldn't be in this situation in the first place, which is why I wouldn't be running Google, and so on. But in every case where there was evidence of someone expressly working around a hold, like in that quoted chat, I'd fire the instigator. No appeal, no exceptions. I wouldn't want criminals working for me, and I wouldn't want idiots working for me either.
US police have run nearly 1M Clearview AI searches, says founder
Re: 99.6% accuracy sounds good
In all US jurisdictions both sides can reject a certain number of jurors without showing cause by peremptory challenge, and others can be rejected by counsel or judge during voir dire.
While the right to a jury trial is a critical civil right in the abstract, in practice it often doesn't work out well. Potential and actual jurors often don't take their responsibilities seriously. When they do serve they often arrive at prima facie wrong conclusions, for the most vapid and hateful of reasons. Unfortunately it's a right that people are often better off not exercising.
Re: 99.6% accuracy sounds good
Have you served on a jury in a criminal case? I have (in fact I was the foreman). There's a lot of pressure on a dissenting minority to agree with the majority during deliberation. In my case I tried to mitigate that as much as possible, and under the circumstances there wasn't a lot of disagreement anyway; but in general all it would take is one or two bullying blowhards to make many jurors decide it's not worth swimming against the tide.
Statistically, in the US, criminal defendants are better off not requesting a jury trial.
And historically, dubious forensic evidence like Clearview's snake oil has frequently been used to secure a conviction. Many judges will disallow the defense from challenging the "science"; this has happened with everything from fingerprint uniqueness to DNA matching to facial reconstruction to bite identification (which is pure, unadulterated bullshit).
Clearview is dangerous and invasive and should be shut down, full stop. No number of legitimate case closures justifies its existence.
Microsoft promises it's made Teams less confusing and resource hungry
Re: Basic UX problems
God, Teams is such a parade of UX agony. A few off the top of my head:
- No option to turn off the fucking "turn parenthesized words into goddamned emoji because all users are pathetic idiots who can't express themselves in text".
- At some point, some idiot decided that all text in a quotation block in chat should be in italics. You can't turn that off. Whoever made that decision should be shot, not just fired.
- God forbid you accidentally mouse over someone's name or (completely unnecessary) "avatar" image during a meeting, because then Teams will show the stupid user-info popup, which can take several seconds to display and then disappear again, hiding a portion of what just might be useful information.
- It's been said many times, but: Teams still lacks a decent search function. And that's despite having at least two of them: the Find box, and the Find box when you enter Ctrl-F, which for some reason does a completely different thing.
- Notifying me every time I go into a chat of who's not around. I don't care, Teams, and if I did, I could find out. As people have ever since computer chat was invented. I do not need you to tattle on my coworkers, or spy on me.
- Oh, I can't tell Teams not to spy on me.
- Downloads always go to the default location. There's no option to make it ask me where I want to save downloaded files, as every other damn application manages to do.
- There's no option to sort the list of Teams you belong to alphabetically. You can rearrange them manually, but only by dragging and dropping, which works poorly and is a huge pain in the ass.
- Oh, and "Teams" the application manages a bunch of "Teams" objects, which contain "Channels" and "Tabs" and all manner of other crap. It's like someone at Microsoft looked at various information models and said "I don't understand what any of this is for!", then drooled on the keyboard for a year. It's aggressively hostile to any sort of sensible organization. Teams is a librarian's dream of Hell.
- Some months ago, Teams stopped ringing my laptop's speaker when a call comes in. I've turned the relevant setting on and off again. I've stopped and restarted Teams. I've rebooted (or, more precisely, Microsoft has arbitrarily rebooted for me). No dice. Another painfully simple, obvious, necessary feature broken.
Teams is the least-competent commercial software I currently use. Maybe that I've ever used.
Is Neuralink ready for human brain implants? Allegedly so
The most bizarre online replacement items in your delivered shopping?
Utah outlaws kids' social media addiction, sets digital curfew
China's best selling smartwatch offers surveillance-as-a-service … for kids
ChatGPT, how did you get here? It was a long journey through open source AI
There are plenty of LLMs already available. Just take LLaMa; it's far more efficient anyway. If you think you really need orders of magnitude more parameters, scaling the model up is just a matter of finding the hardware resources. There are plenty of corpora available to feed your shoggoth.
While I sympathize with the article's tone, the argument's over table scraps. Here's the thing. Building an LLM is easy. Training it is easy, if expensive in hardware costs. LLaMa has shown you can easily filch another model's "personality" by fine-tuning against it, so you can make your own model take on the surface appearance of GPT-4 if you really want to. That means there's no competitive advantage in the model itself, aside from the hardware costs of running the thing, and once someone's made a Really Really Big model you can in effect steal much of its "feel" using a model that's merely Really Big.
So why are Alphabet and Microsoft and Meta1 fighting over this? Because the scarcity, and thus the value, is in context. Alphabet owns a ton of user context, in search history and the web index itself Gmail and Google Docs. Microsoft owns a ton of user context in 365 and GitHub and Bing's web index. Meta owns a ton of user history in cat pics and racist rants, which isn't as rich with business data and the like but is good for manipulating people, so it's pretty valuable. That is what makes their control over LLMs valuable. And that's what other people don't have, and why the big LLM owners will crush the zillions of startups.
LLMs are going to shift even more attention and power to the big players, because while the typical Reg commentator might hold his or her nose, most people will happily stroll into the information abattoir for the convenience. Even a large proportion of techies; we're already hearing all this noise about how LLMs make programmers so much more productive, and how it's the future of software development – as if writing new code weren't the single least important thing a good software developer does. (If you write enough code for an LLM to significantly improve your productivity, you're almost certainly Doing It Wrong.)
1There are other players, but the Chinese ones are effectively in a different market, and I don't think any of the smaller ones have a chance.
Page:
- ← Prev
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 20
- 21
- 22
- 23
- 24
- 25
- 26
- 27
- 28
- 29
- 30
- 31
- 32
- 33
- 34
- 35
- 36
- 37
- 38
- 39
- 40
- 41
- 42
- 43
- 44
- 45
- 46
- 47
- 48
- 49
- 50
- 51
- 52
- 53
- 54
- 55
- 56
- 57
- 58
- 59
- 60
- 61
- 62
- 63
- 64
- 65
- 66
- 67
- 68
- 69
- 70
- 71
- 72
- 73
- 74
- 75
- 76
- 77
- 78
- 79
- 80
- 81
- 82
- 83
- 84
- 85
- 86
- 87
- 88
- 89
- 90
- 91
- 92
- 93
- 94
- 95
- 96
- 97
- 98
- 99
- 100
- 101
- 102
- 103
- 104
- 105
- 106
- 107
- 108
- 109
- 110
- 111
- 112
- 113
- 114
- 115
- 116
- 117
- 118
- 119
- 120
- 121
- 122
- 123
- 124
- 125
- 126
- 127
- 128
- 129
- 130
- 131
- 132
- 133
- 134
- 135
- 136
- 137
- 138
- 139
- 140
- 141
- 142
- 143
- 144
- 145
- 146
- 147
- 148
- 149
- 150
- 151
- 152
- 153
- 154
- 155
- 156
- 157
- 158
- 159
- 160
- 161
- 162
- 163
- 164
- 165
- 166
- 167
- 168
- 169
- 170
- 171
- 172
- 173
- 174
- 175
- 176
- 177
- 178
- 179
- 180
- 181
- 182
- 183
- 184
- 185
- 186
- 187
- 188
- 189
- 190
- 191
- 192
- 193
- 194
- 195
- 196
- 197
- 198
- 199
- 200
- 201
- 202
- 203
- 204
- 205
- 206
- 207
- 208
- 209
- 210
- 211
- 212
- 213
- 214
- 215
- 216
- 217
- 218
- 219
- 220
- 221
- 222
- 223
- 224
- 225
- 226
- 227
- 228
- 229
- 230
- 231
- 232
- 233
- 234
- 235
- 236
- 237
- 238
- 239
- 240
- 241
- 242
- 243
- 244
- 245
- 246
- 247
- Next →