I think you could also make the case that the existing abstractions aren't actually fully deterministic themselves. The compiler or interpreter may not behave as it should. Therefore, for any correct C code, there's probability that the GCC compiler will turn it into correctly formed machine code. But it may not!
Is the probability much higher with GCC? Sure. But it's still a probability.
I am sorry but this is an insane take. The probability of GCC going haywire with your special snowflake correct C code? Please. Have this EVER happen to you? I am not talking about the performance of the generated assembly because that IS flaky, but functionality wise I do not think so.
If people are so confident about the determinism of LLMs, or at least consider it on par with compilers, please ask it to compile your source code instead. Better yet, replace all your GNU utils with LLM instead. Replace your `ls` with `codex "prompt"`.
I have done this, alias codex --yolo -p . It's very helpful not having to remember every odd command and its parameters. It's a bit more typing but I type faster than invoke and scan through man pages.
Non-deterministic systems produce different output states given identical input states.
Even if a compiler's memory gets a one-in-a-million bitflip that produces a different output, it doesn't mean it's non-deterministic. It just means that the output state is different due to an external force changing the internal state.
An infinite loop will halt when the processor is powered off.
yes, because 30 days had passed from the time the patch landed in the kernel, as per industry standard.
approximately every security researcher, including the likes of google and other big names you may know, does a 90+30 disclosure, which is what happened here. they do this for good reason, which has been figured out over decades of experience in reporting thousands and thousands of vulnerabilities.
the only security researchers i know of that dont like 90+30 actually argue for shorter timelines (or immediate disclosures).
What do you think went differently in this case versus other high profile vulnerabilities that had binaries already available for major distros? I feel like it often (usually?) works out that major distros have kernel packages incorporating the fixes already available.
Is this just down to luck, a quirk in the timing about when Linus merged the fix versus when the release gets cut?
The trouble with shrdlu and all the other early "natural language" parsers is that they weren't really parsing natural language. They parsed a formal language with syntax designed to look kinda like English.
At the time it was believed (by Chomsky, etc) that natural language could be described in formal terms and parsed by fixed rules. But after several decades of failed parsers (and the success of statistical methods like LLMs), it is clear that formal and natural are fundamentally different types of languages.
When I was in college (back in the early 2000s) I took a course on computational linguistics as part of a linguistics degree. And we studied formal language style parsers, and statistical style parsers. And I really didn't like the statistical style parsing—coming from a math background, the formal language style parsing seemed a lot more elegant to me. But damn if the statistical parsing didn't work a lot better :-(
You're missing a step. The perturbations are not fully random. The LLM also looks at the result and tries to do credit assignment to determine what changes to try in the next round.
Other professional critics like Gary Marcus and Emily Bender are the same way. It doesn't matter what neural networks do, they will always be a dead end that should be abandoned.
The downside is that since everyone can do it, without understanding the "it" properly, security issues will be boundless and not understood, being rooted will be commonplace, and what you thought was safe and secure will be widly broadcast.
For most people, though, prompting your own software is beyond the realm, since they have day jobs to attend to, groceries to buy, children to herd, and lawns to mow, and they will be oblivious to the scams, fakes, and charlatans who have vibed up something to look useful but only aimed at getting hold of personal info and credit card details.
>security issues will be boundless and not understood
Again with the optimistic take, but I do not think this will be an intractable problem. LLMs are becoming good at finding security vulnerabilities.
This would certainly be a radical change in how the software ecosystem operates. But I think you are ignoring the advantages of more flexible, abundant, customized software.
sorry you mean the upside is that selling software is dead? I really find it optimistic to think everyone even whats to "prompt" their own software. That seems like a nightmare compared to quality software developed by experts that you can trust instead of inshallah and AI
> I think we are seeing the beginning of the end of for-sale software.
We're seeing the end of "simple" for-sale software. Like OPs CRUD app, a UI front-end on-top of a database, of which there are a gazillion examples so some AI can easily synthesize some approximation of whatever requested variation.
The selling of software was always in the "moat", not how fast you were able to churn out CRUD apps. We used offshore that to a more viable economy, but now we're offshoring that to an automated process.
We're not seeing the end of for-sale software, we're seeing the beginning of the end-to-end solo founder.
I posted more or less the same thing in a comment over on lobste.rs[1] - being able to create your own bespoke software tools, without any developer experience is (mostly) a really cool thing.
This isn't someone being inspired to build something: It's the automated "drive-by" cloning and scammy, dubious nature of these clones that bothers me along with the copying of personas & identities to spam them across social media.
we're already there. When apple announced the iphone, the price of software went from can-make-a-profit-$39.99 on a PC to $4.99 and then $0.99 for a phone app. Which became "that's too expensive" at some point, and then reset to free.
I don’t have data on this. But I’m getting recommended YouTube videos that are 1-2 hr AI generated music, in the genre of background music (coffee listening and focus).
I listened to one. It was pretty good!! There’s no lyrical content, but the production was strong.
In that niche of “music you don’t really pay attention to” I predict AI generated music will only grow.
That’s basically the face of GMOs, so it is an issue for GMOs. GMOs for whatever reason have a terrible ambassador and I haven’t seen evidence to the contrary.
For vaccines, a good portion of the population remember vaccines being developed and marketed to help people. Then there are immigrants that remember more recently how life changing vaccines are.
I am very concerned by the rise of political violence in the US, and I especially don't like how much support it gets on social media. Burning down a warehouse or shooting a politician does not make you a hero.
What's the alternative? You think calmly asking those politicians not to sell you out to the trillion dollar corporation that wants to build a datacenter in your backyard is ever going to work? Be real.
History has repeatedly taught us that violence is usually the answer. I wish it didn't have to be this way, but it is what it is.
I think it’s interesting that you choose to focus on this part of the situation. To me, it’s far more relevant that the general public has little, if any, recourse through legal means such as voting. This is what makes political violence inevitable, and some would say, fully justified.
Political polarization create tribalism, where people align their view with their tribe, and justify an increasingly more escalatory means to fight the "other side".
Other potential macro-contributing factors may include: breakdown in local community, removal of community forums for discussion, attention economy and tabloid journalism gravitating toward emotional reaction (TikTok) rather than intellectual dialogue (balanced journalism), social media echo chambers, removal of accessible popular education, defunding of public media, unaffordable public access to medicine, credit culture, increasingly unaffordable costs of living and abnormally performative political dioramas. The net result are people, unable to reason about the world around them, drawn in to emotional us-and-them with a dialogue of echo-chamber reinforcement, who decide semi-rationally to "chuck it all in" the second things get out of control financially, psychologically or emotionally. In other words, the modern world has built a perfect breeding ground for recruitment to extremism. <s>Great time to start a cult.</s>
... and in a classic example, apparently the mere mention of concern regarding the rise in US political violence got this thread flagged. Where can you have a discussion anymore?
Especially consider how many fellow workers Paper Mario could have killed with his arson. But smart people tend to realize they can do more with their lives by not being violent.
In my book, you are a hero if you sacrifice your own well-being for the utilitarian good of the public.
Many people here would call Putin's assassin a hero, the important distinguishing factor is whether it's a clear societal good or bad. If it's unclear then it's assumed bad.
I am not disagreeing with you here. But platitudes do nothing to convince people. You need to actually explain why the world is a better place with X politician in it, because it does actually matter.
Violence isn't going to give you the quick answer you think it will.
Once you start shooting, everyone starts shooting. Bystanders get hit. Companies start defending their businesses with private armies. The economy collapses. We all lose.
Countries high in political violence are the worst places in the world to live.
And LLMs can handle very abstract concepts that could not possibly be encoded in C++, like the user's goal in using software.
reply