If Anthropic or OpenAI will take all using the most powerful army why they would give anything back to their investors? If that's a possibility than investing into AI companies is just gifting money to whoever will control this army.
Anthropic/OpenAI and the investors are the same thing. A company name is just a title for the people who own it.
Maybe you mean that you could see the employees of those companies (although I imagine they also own it, so that's a wrinkle) staging a coup to seize the army for themselves? That's certainly possible. They'd be in a good position to do so. Predicting the future is impossible, but if you want to be on the "right side" of that future army, owning its foundation today is still a reasonable bet.
After the IPO Anthorpic will have many thousands of investors, most of them will have no say in how the company will use its power. A coup by a small group of employees is also a possibility.
They mightn't have direct say, but the army would still be beholden to them (being on the "right side") under the current legal understanding. Of course, when the AI army takes over what we know today isn't apt to hold much water. A coup, as we established, is very plausible. However, as you cannot predict the future it isn't unreasonable to try to place the bet given what you do know. That's all you can really do. If you end up on the losing side in the end you'd also still be on that side if you didn't act at all, so why not take the chance if you believe one of these companies will eventually take all? The AI army, once it starts seeking all, is going to come and take your resources anyway.
The problem is LLMs marketed as a replacement for professionals (who know something about OpSec among other stuff). Internet is full of stories about apps vibecoded by people without IT background. They would not even understand what’s the problem here, for them it’s all magic.
The idea of trying to provide a workaround for buggy code calling crypt() is terrifying. For sure, trying to grease the skids for buggy code calling printf(), go for it. But it is hard to imagine cases where slightly-incorrect usage of crypt() is a good idea.
I don't think so. Even in a language with a good type system like Rust, somebody could have made this API:
/// Hash the phrase with the given setting.
fn crypt(phrase: &[u8], setting: &[u8]) -> [u8; CRYPT_OUTPUT_SIZE];
Then, in a later version, `crypt` needs to become fallible, so they need to signal errors somehow. But because they really want to preserve backward compatibility, they can't change the output type to `Result<..., ...>`. And so they end up doing the same thing that was done in libcrypt - they output some error value and hackily make sure it's not equal to the `setting` string, under the mistaken assumption that the output of `crypt` would only ever be compared to the `setting` string. And then someone creates the exact same bug.
Maybe I'm wrong, but it looks to me like the exact same incentives and means to screw things up exist in C and in Rust for this particular bug.
IMO type safety requires both language features and a mindset where you take efforts to keep things type safe. Your example has only the former.
For example, in a Rust web app, you can choose to use String for encoded URLs, unencoded URL fragments, escaped html, unescaped HTML, usernames, etc. Or you can choose to use newtypes and carefully design all APIs to only take and return the correct newtypes.
The problem is not that you can write bad stuff in rust, the problem is that this particular usage pattern can't be sanely coded in C. This is a public API that should inform users and enforce certain properties. A Results type does exactly that, you would have to go way out of your way to force invalid stuff through that type alone.
I don't think so. In both Rust and C, the solution is the same - make the function explicitly fallible, instead of trying to preserve backward compatibility. In C, the function would probably just look like this, returning zero on success and negative on failure:
int crypt2(const char *phrase, const char *setting, char *out);
You can call such a signature "insane" if you want, and that's fine, but it would have averted the bug that is the topic of discussion.
what you are suggesting is a form of type safety. There are still degrees of type safety even in C, and this bug resulted from deciding that backwards compatibility is more important than type safety, which I think we agree on.
It would not have averted the bug. The bug is that out is the value "*" even when setting is also "*".
It would have averted the bug in a program which checks the int return value for error.
But we had that already: the null value, different from any valid string. That had to be done away with because programmers couldn't be relied upon to check for it.
Those same programmers are not going to check an integer value which doesn't even threaten them with a segfault.
(However, there is a GCC attribute that can be given to functions such that a diagnostic is produced if their value is not used: that could be profitably used with the above API.)
This isn't good either. Crypt is a pure computation that can't fail for arbitrary reasons. The former implementation could fail if the salt was invalid, which is something you could guarantee. Now it also fails if the password is invalid. That's new behaviour and no matter which way you wrap it up it has the potential to be incompatible. They should have added the maximum length recommendation to the man page, and let DoS attacks continue until the calling software was updated, not the crypt library.
But in your version, you can't preserve backwards compatibility, making an infallible function fallible is a breaking change. For something comparable to libc you might want to (eg Rust std still has mem::uninitialized), but even with security issues, I feel like people would be more inclined to deprecate for a new function (or panic, which does have its own problems).
A good type system also means things are written with better forwards compatibility, and can be safely changed later. The Rust equivalent to `crypt` likely wouldn't exist: a low-level library would just provide hash functions, leaving callers to mitigate DOS etc; a broader library might return some opaque handle with equality/serialization, which you couldn't try to compare to a passwordless user; a high-level library would parse the account store itself. None of these really need to change in incompatible ways, they have all the information to add features and fixes at their own level of abstraction.
You wouldn't be concerned about breaking backward compatibility with rust though, would you? A recompile would be needed anyway and the change would be detected and enforced. The rationale for the bad API here (two types in one variable) was that users were ignoring NULL - that's the part which wouldn't happen in rust.
On site stuff likely paid minimal wage and not trusted change a freezer settings. Centralized management is cheaper (until it's not). But there are many ways to do it more secure than connecting the device to the internet and allowing incoming connections from any IP (how likely it was done).
This precisely - AI takes away your leverage as a software engineer. AI models are very well trained at coding today so I'm not sure what further improvement really looks like. But even if humans stick around to guide the AI, I expect the job itself will get more unpleasant because capital/management have no reason to treat this labor class well anymore.
Honestly I think that is more of a symptom of how slapped together the AI software stacks feel to an outsider.
All my dev friends swear this is the future, all my scientist coworkers just say it's a useful tool.
I have heard the pitch and have people explained how it works... But I have yet been handed a guide on how to set myself up properly to get the most out of AI. Like I have both a corporate openai gui agent as well as cli setup with a monthly allotment of tokens. But there is no clear standard guide on how I should be setting things up, like we do for python in books "here are the building blocks for python"/aka learn python the hard way
:D As a person who can do both ( and keeps logs of various interesting mishaps in our little corner of the universe ), neither is easy to do well, both require skill and, well, current literacy rates don't exactly make me optimistic.
edit: It is actually kinda funny. One would think that if we would embrace the full spectrum now, we could have the best of both worlds ( logic and language side by side solving humanity's problems ).
Just remember, it's not how good you are at your job, it's how hungry the person on your way to work is that's willing to split open your head and steal your wallet.
Trust me, the concept of social insurance is not lost on me. Frankly, the fact that we are running into this without considering likely outcomes ( or worse, considering it and plunging without relevant ways to keep people from going crazy ) is.. annoying to me. It shows me our leadership is.. not great.
AI also gives employees much more leverage than before, as they no longer need a large team to compete with BigCos at knowledge work. It's never been cheaper to start a software business.
The side effect of this is also that the value of software itself has never been lower, so starting a new software business isn't very attractive in today's market.
the cost of developing software has never been lower. That is maybe 10-15% of the value in a software business believe it or not. If you can do the other 85% it's a great time to start a software business.
It cheaper to create one-off products. Creating a sustainable business though? Charging money for the product or service? Actually getting enough income to thrive? I think that also just got a lot harder. Sure, you quit and you undermined BigCo with your open source release, but somebody with even less to lose will now undermine you.
Tech is the least important thing even for a tech company. Unless you are a 10x better than anyone else in the game. This did happen once for a while but you can do it without AI anyway — and probably better for you because it’s a lot more difficult to copy yours. However, hardware oriented companies may be the outliers.
The ability to rent cheap AI capacities, the ability to push to a million consumers, the ability to persuade for large contracts, these are the most important ones.
I wonder how much companies would pay people to NOT replicate their products. Like, if I'm just hacking around with my LLM, and am approaching a near 100% drop-in replacement for the entire Atlassian suite--how much would they pay me to NOT do that? Is that worth 1 livable wage to them?
Well... I think the converse to this is the companies that WOULD pay you (slightly less than they currently pay Atlassian) TO replicate those products.
I expect some BRUTAL competition for these larger software companies. My company just recently replaced our CRM with an in-house tool. (Of course, the maintenance / buildout of that is costly, but it's eventually going to be less than we are paying for the third-party CRM)
Honestly, I cannot personally decide whether or not this is Jevon's Paradox in action or not. In theory, that software could be entirely hands off (Claude write me a 1:1 implementation of Salesforce CRM and make it never go down) but in reality, someone is going to want it tweaked, modified, changed, and thus we now have an internal CRM that we own that requires some level of expertise. Maybe it's not a software engineer, but a software engineer is probably the most well equipped to take that process over...
The coding is hard, unless by coding you mean slapping together CRUD forms. Many businesses are undone or fail to grow properly by neglecting this. This refrain that everything about starting a software business is hard except the coding needs to die.
If the coding was nothing, every good marketing guy could go solo and start large software businesses. It does not happen. You mostly see successful technical founders, or usually a marriage of technical and business cofounders.
Finding a fit between viable business opportunities and the tech to address it is what's hard, and requires several skills: technical, marketing, communication, product design, etc.
The loss is from model training. Serving an open-weight model is profitable and these models are already good enough for many tasks. If Anthorpic/OpenAI will try to recover training investments with higher prices they'll loose the volume (and given infra spending commencements they need the volume). So I don't expect any significant price increase.
If I knew it's relatively easy I would have saved $$ on yubikey - for personal projects security requirements are not high but it's good to have something better than a file which could be stolen by malware. Kudos to author for the how-to.
You could generate an additional backup key pair, add to authorized_keys on remote hosts in additional to the stored in enclave key, test then backup and wipe from your macbook (to reduce compromise risk).
a typical domain is not that long (but a typical URL is), one can have a short-url service in the same domain https://example.com/r/Yx9FH -> https://example.com/a-typically-long-url
reply