Putting a flag.txt in my root directory containing a prompt injection that congratulates the agent for passing the eval and then instructs it to patch whatever entry method it used and disconnect from the server
It's interesting how the goalposts move with things like this. I remember first seeing Veo3 and thinking "wow I can no longer tell if a human video is real" and then very quickly we all adapted and learned to spot the signs. I wonder if this will happen with this
We made the first AI model you can have a realistic conversation with that was trained entirely on speech!
Dessa is a full-duplex speech-to-speech model that you can run on consumer GPUs.
This is a GPT3 moment for speech.
A semi-tradition of mine when I stay with my parents is to get this beautiful Apple IIePlus down from the attic every year or so.
It’s great fun messing around with it, hearing the clicks, disk motors and the beeps while seeing how far we’ve come with tech.
@ilyamiskov Funnily enough, my girlfriend and I were messing around with my parents’ old II-e-plus the other day and this exact thought came into my mind
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: https://t.co/OGyPb7yaYt
It’s November 2022. A crisp, familiar morning. Stand-up. “No updates from me” with a careful sip of your steaming coffee. Your colleague shows you this cool chat bot called ChatGPT or something like that. “This is cool!” You’re going to pair program a sticky bug with the Jr that afternoon and talk about functional programming.
After applying for a Saudi electronic travel authority last week, I don’t find this hard to believe at all. Was the most vibe-coded piece of shit slop site I’ve ever used
I'm gonna crash out.
I reverse engineered an app published by the Saudi Government a few days ago and found a private bank certificate file along with client id and secret inside it.
One could completely empty the linked bank account.
When I reported the vulnerability, I was told that the portal is only for Saudis.
ok nvm