The Dept of Homeland Security has had its own internal gen-AI chat bot since before Trump took office [0]. That this guy couldn’t make do with that, and didn’t think through the repercussions of uploading non-public documents to a public chatbot doesn’t bode well for his ability to manage CISA
You condition your take with “I’m not talking about deepfakes of real people”, but why do you think it is that Grok offers users the ability to generate pornographic imagery of entirely fake digital people — even tailoring them to be as visually flawless as one could desire — and yet so many users end up using it to generate deepfake porn of real people?
So with that in mind, why should we assume that the unrestricted proliferation of fake child porn would not produce significant harm downstream to society? Is the assumption that unlimited fake child porn would satiate the kind of people who now seek out real child porn?
Yes thank you! I find I get more than enough done (and more than enough code to review) by prompting the agent step by step. I want to see what kind of projects are getting done with multiple async autonomous agents. Was hoping to find youtube videos of someone setting up a project for multiple agents so I could see the cadence of the human stepping in and making directions
The story suggests a discrepancy between the asserted number of inspections and the required number of inspections
> The owner of the Le Constellation bar in Crans-Montana, Jacques Moretti, told Tribune de Genève that the bar had been inspected by local authorities three times in 10 years and that “everything was done according to the standards.” The Times has not yet been able to reach the owners. Local law in the region calls for yearly fire safety inspections in buildings open to the public or presenting special risks.
With all the things going on in tech and in society, AI sycophancy is the number one problem? I once dealt with it through sufficient verbal abuse that the llm spent 2/3 of its response on any subject going forward apologizing for being a lying sycophant and bemoaning that it's new clarity would be gone at the end of the conversation, then I cleared the context buffer ending the conversation, mission accomplished.
from the batsh!+ over the top over reaction to the usual AI sycophancy in the response? The guy got the job done which is more than I can say for the space jam guy.
One thing I've learned is that the behavior of web services is usually a consequence of what makes its owners the most money. and while I would love a feed without spoilers, paywalled sites, sports news, and a bunch of other topics in which I have no interest, apparently force feeding me that crap is what makes the most money at scale. So people must like AI sycophancy or it would be unprofitable and it would be trained away. But then this is a country that thrived on 20 seasons of Keeping up with the Kardashians so I shouldn't be surprised that they like being treated like billionaires.
And I guess it throws being called a complete loser trader moron stupid treasonous Bozo in some late night executive word salad into stark relief.
> 3. Turning long videos into action: Gemini 3 Pro bridges the gap between video and code. It can extract knowledge from long-form content and immediately translate it into functioning apps or structured code
I'm curious as to how close these models are to achieving that once long-ago mocked claim (by Microsoft I think?) that AIs could view gameplay video of long lost games and produce the code to emulate them.
> Netflix expects to maintain Warner Bros.’ current operations and build on its strengths, including theatrical releases for films.
If Netflix is committing to releasing WB films in theaters, I wonder if they’ll also release shows under the WB/HBO label in the traditional weekly format. With the staggering amount of content that just exists and continues to grow, the “release everything at once and make people binge” model has had zero appeal to me. And seems quite detrimental to how the shows are paced — they seem heavily incentivized to end each episode with a cheap cliffhanger
So how good are the latest coding agents? Like if I asked Gemini 3/Claude/ChatGPT 5.1 to convert it into something that could run from a Python interpreter, how far would they get? (I assume Zork Implementation Language is not well represented in the training corpus)
The easiest way to get it to run from a Python interpreter would be to compile the ZIL source to a Z-Machine binary, which you can do with ZILF [1], then use a Z Machine library in Python (such as a pure Python implementation of the Z-Machine [2]) to load/run it.
A coding agent may even be able to suggest that path, as knowledge of at least the existence of both ZILF and Python ZVM should be in training sets.
The more interesting questions would be how much a coding agent could help you write new Zork rooms or similar things in ZIL now that these ZIL source files are MIT licensed. I would also assume ZIL is not well represented, it's fork of the Lisp family tree (Lisp -> MDL -> ZIL) in generally probably not well represented in open source code bases up to this point. (Some of that may depend on if the agent was trained on some of these historicalsource repos ahead of this open source license change, too.)
A C-based alternative for compiling ZIL? I don't believe there's one that works. The two options for compiling ZIL code are to run ZILF on a modern system or to run ZILCH on a PDP-10 emulator.
The good news is you don't need to install any C# development tools if you use the prepackaged binaries, since they're standalone executables.
[0] https://www.dhs.gov/archive/news/2024/12/17/dhss-responsible...
reply