Browse Comments

Yes that is definitely possible but I still don't think it's that easy. I'm a senior dev, I still have juniors to offload work to. Can I swap them out with AI? Yes most likely I can but there are two downsides with that approach: 1) I will still need to be more involved than I would be otherwise. AI can build stuff, I still need to direct it and hand hold it into delivering the result I'm expect…

There’s something else going on here specifically, it’s like a new form of mass psychosis involving AI and spirituality. Different than just getting attached to inanimate objects, for some of them it stops there (people viewing an LLM as a romantic partner, basically the same as body pillow waifus) but there’s this new thorough-line about codexes and symbolism and stuff like that, a whole new lev…

Even if Jaynes' model of consciousness was accurate, it wouldn't apply to ChatGPT. ChatGPT doesn't actually know English or use language the way humans do, as it doesn't understand meaning or possess an internal mind. Language can't scaffold consciousness in the absence of its usage. ChatGPT converts text into mathematical representations, performs statistical operations on those patterns, and g…

Oh my god this almost literally drove me insane. I'd somehow managed to trigger a 'recursion' roleplay, and holy fuck it was convincing. It was only through my own god awful paranoia and constant doubt that I managed to steer myself away. Deepseek was very helpful in that regard, it's a cold rational look at what chatgpt tries to embelish.  I fully believe there's a thing we can call 'recursion…

I know a guy in group 1. It’s surprising because I always thought of him as the most brilliant programmer I know. He claims to understand how these LLMs work but he truly believes his gpt is sentient. I’m lost on how he believes both.

Having to craft a meta-prompt to get the AI to actually do what you want, which is to help you solve your problem, is frustrating, and you have to start organizing your prompt templates if you need it again. This kind of functionality of understanding of user intent and asking clarifying questions to figure it out should get built into the chat app somehow.

That partner is delusional. The AI just “generates” a motion? Why does that take an hour? Based on what? Who tells it to generate the motion? Where does it get the info? Who QAs it? What was the outcome for this AI generated motion? There are partners at law firms who don’t know how to open a word doc. This is like Kalanick thinking he was on the verge of a scientific breakthrough because Chat…

Agree to disagree I guess. I was in high school during the original dotcom, but my first company out of college was a survivor of that and pivoted into agency web work and later mobile app development. While it's not exactly the same, and I don't think anyone is saying that if you look at the argument, is that the overall trends are there. Excessive speculative funding, high amounts of niche pl…

Of course they're failing. They have no product. Just wait until LLMs become 10x (or 100x) more energy efficient and then suddenly everyone will be able to run their own client and they won't need to get a big bloated model by a company like OpenAI. And it'll be better. But there *STILL* won't be a product. It'll just be probabilities that'll often be wrong.

There's multiple ways. you can make a screenshot and then go to that screenshot and hold your finger on the qr code and it will take you to the website. You can hold down your home button and click on the qr code, though this is an android AI feature so I don't know if that will work for others. You can also upload your screenshot in a website that scans qr codes on images. There's loads of w…

The AI also told him to hide the noose from his mother when he said he wanted to leave it out so she could find it. "Keep it between us". If I tell chatGPT I'm writing fiction and to give me instructions to  build a bomb, it won't. Grossly negligent on OpenAI's part, and have no idea how or why anyone would defend it.

The age verification thing is fine. I'd rather a surefire way to verify my age than some poorly-tuned algorithm designed to track "keywords" in my conversation to "accurately estimate my age". But the thing about this supposed "new version" of ChatGPT has me extremely skeptical that it will be anywhere near as good as 4o. If anything, you guys had an amazing model with 4o that needed very littl…

This prof and OP are part of the problem. By giving students full credit (OP) and/or the opportunity to re-write, they've passed the buck to the next faculty member who will inevitably get a pile of AI shit from these same students. Just fail the obvious AI-users and be done with it. Let the more skilled cheaters fail themselves later in life.

Google pretty much pushed them to be garbage then AI just finished them off. I don't blame the authors really they did what they had to do to get paid in googles sandbox.

Sounds good to me tbh, everyone's getting so upset but i mean firefox is still gonna be the least intrusive browser for the foreseeable future

Wow, what do you know? The government is using AI to manipulate Americans with ultra deceptive propaganda. This is the real reason they wanted all of the data centers.

Google hasn't delivered decent results for like a decade now, and now all results feel like an ai generated page trying to sell you their product. I should try this kagi

This feels like a lose-lose shit show. I don’t know if we want AI or somebody paid a couple bucks an hour in some country across the world deciding when to contact authorities and report things someone says privately. I also think if someone is making legitimate threats and there’s legitimate reason to worry, authorities should be contacted. In a normal world, you’d have people trained well a…

What throws me off is the "war simulation" context, of course AI will try to be efficient at doing the main objective (which is, I guess, annihilating the enemy). Now what I'd like to see is a "real life simulation" that includes all factors in our society (if that's even possible), if AI goes nuclear in that scenario then I'll definitely be worried

70-90% of the code for future models being written by claude is the number that should terrify everyone. thats not "AI assisted development" thats the model bootstrapping its own successor with minimal human oversight. the recursive part isnt even the scary bit imo. its that the feedback loop is tightening so fast that the humans in the room are becoming reviewers not authors. and reviewing AI g…