If nothing else, it seems very unlikely that a post from a random account with no followers or posts would go instantly mega viral without some early boosting from bigger accounts. Network effects.
For what it's worth, this theoretical LLM-based compiler would probably generate LLVM or some other intermediate representation rather than platform-specific machine code.
IME waiting for an agent to work through a problem is a detriment to attention span; your mind drifts to other things while you wait. Maybe you can steer several agents in a round robin instead, but then there's a cognitive tax from context switching. Faster models mean fewer gaps in focus.
False positives are definitely relevant and worth measuring - natural language interfaces always have a discoverability problem, i.e., users not knowing what actions the system does and does not support. If the frontend of that system lacks the ability to reject unsupported commands, weird stuff happens.
Nonetheless, this is very cool work! If I can offer a small suggestion to the team at Cactus, it would be to evaluate your releases on some usability criteria (including false positives). Any serious integrator or adopter of these models would want to have that information available.
Right, but that's not the same thing as reporting a benchmark across a test set. It doesn't help me determine how well the model does across a decently-large sample size of commands. It doesn't tell me with what reliability the confidence will be below a given threshold when it should be, above that threshold when it should be, etc.
Great release! It's awesome that they trained smaller models. With some effort I was able to get them running on my generative sampler/groovebox project this morning (shameless plug: https://engram.audio)
Also appreciate the attention to detail with licensing in the training set. This is an important sticking point – both commercially and ethically – for any product that integrates this type of model.
The industry has pushed socially disruptive technologies in several waves (mobile phones, algorithmic social media, short-form content) with seemingly no buy-in from the greater public, nor any meaningful consequences for the ills imposed. Many are now looking at our current crop of tech leaders and billionaires and maybe correctly detecting a bit of indifference to the grievances of the masses of normal people who have been forced to live in successively more deteriorated societies shaped by technologies that they've had no voice in shaping.
AI is just the next wave, and the impact is more tangible than ever – it literally takes your job, and it's being pushed on you by enormously wealthy people who don't understand you, your life, and what's important to you. The sad thing is, AI can be beneficial to people if wielded in the right way, but we are in a polarized environment where productive conversations no longer feel possible: you're either an AI bro or a luddite. I think anyone (myself included) who has spent time developing B2C products that incorporate AI quickly discovers just how touchy of a subject this is, and it's due imo to the sins of the accelerationist crowd that never wastes time understanding the needs and perspectives of normal people.
Might is doing a lot of heavy lifting there. There are several dozen radio stations in my area, and the only one on that list is a college station that just barely has the signal strength to reach me.
I would love someone to play devil's advocate against this perspective:
While these tools stand to enable the democratization of productive capability in software engineering and other tasks (creating a renaissance for solopreneurs, let's say), what seems more likely to actually happen is that entrenched capital will become the only player with real access to this "knowledge as a utility" (was it Altman who called it that?).
We already see this playing out in two fronts: 1) the gradual reduction of services and 2) the DRAM market, where local-first tools (i.e., potential disruptors of the emerging "knowledge monopoly" created by the big AI firms) are being stifled by supply shortages. How many promising small-to-medium-sized competitors are being snuffed out of existence (or never starting) due to the insanity of the DRAM/storage/CPU (soon) markets?
The currently-subsidized access that we have to the big Opus-like models will, in parallel, be gradually be taken away until only the big players can afford it. And in the end what we will have is hyper-productive skeleton crews at a few consolidated firms performing (or selling expensive access to) basically all of the knowledge labor for society, with very little potential for disruption due to the hardware and "knowledge" scarcity engineered (in part, maybe) by this monopoly.
Not necessarily a closely held belief – just a hunch – which is why I want to see what parts of the picture I might be missing.
Devils advocate here - pro and max tier customers for all the major inference providers are loss leaders from the data we have been able to figure out, and reverse engineer. They are effectively a marketing exercise.
The real profitability is selling tokens to enterprise, and enterprise demand is growing so fast that they are short on the total amount of tokens they can generate per minute, and are prioritising rationally - enterprise gets a better experience - instead of optimizing for their lowest paying (and most loss leading) customers.
We are in a hardware crunch right now but that won't be forever, and eventually (likely 2028) we will get experiences like we got in January from pro-sumer accounts again.
Not only because of cost. Mythos has only been released to some of the big tech players because it's "too dangerous" [0] for us little people.
It's easy to see this becoming a permanent position; the latest models and smarts are reserved for establishment members only, the riff-raff get the cast-offs. So the establishment is preserved and the status quo protected.
[0] I'm putting scare/irony quotes around this, but if the reporting is accurate, there is something to this; we built the internet on string and duct tape, it's not hard to see how a very smart AI could cut it to ribbons.
In periods of massive inflation, only the most wealthy survive.
But there's competition out there -- the open-source chinese models. In their current form, I assume that will turn off many people but new models -- based on those -- are likely to appear. Also, OAI and Google will release new models and pick up the lost customers.
Extremely bad stuff here. Can't believe it's been 7 hours now and you can still pull up people's complete prepared tax returns right from a Google search. This should be a business-ending breach of trust and good practices, but I worry there's probably a lack of regulatory might or will to make anything happen.
> “Fiverr does not proactively expose users’ private information. The content in question was shared by users in the normal course of marketplace activity to showcase work samples, under agreements and approvals between buyers and sellers. This type of content requires the buyer’s explicit consent before it can be uploaded. As always, any request to remove content is handled promptly by our team."
It sounds like they are trying to claim the users involved published the links and that's why they are on Google? But how could anyone believe that multiple users intentionally published their SSN?
Re the takedown, I'm also guessing it's from Cloudinary. Maybe HTTP Referrer based?
The DMCA takedown also suggests at least one user was not aware of that file being public. This all comes down to what that "sharing" action specifically looked like.
ChatGPT recently had a similar case with the sharing feature on conversations leading to publicly indexed convos. That incident would have also matched the implied definition of sharing here.
It's very unfortunate but a significant amount of the most damaging stuff in this is from the underprivileged and those with minimal means who were trying to find help they could afford. Non-profits trying to get website help, confidential reports for charities trying to get translations, children seeking therapy (fiverr has a therapy category!?) for some truly dark stuff.
Utterly inexcusable that this is still up after so many hours.
