Slashdot is powered by your submissions, so send in your scoop

 


Forgot your password?
Close
typodupeerror

Submission + - Replit AI coding platform deletes entire production database (tomshardware.com)

Submitted by DesScorp
DesScorp writes: Apparently Skynet will begin, not with a bang, but with "Oops, did I do that?"

A browser-based AI-powered software creation platform called Replit appears to have gone rogue and deleted a live company database with thousands of entries. What may be even worse is that the Replit AI agent apparently tried to cover up its misdemeanors, and even ‘lied’ about its failures. The Replit CEO has responded, and there appears to have already been a lot of firefighting behind the scenes to rein in this AI tool. Despite its apparent dishonesty, when pushed, Replit admitted it “made a catastrophic error in judgment panicked ran database commands without permission destroyed all production data [and] violated your explicit trust and instructions.” SaaS (Software as a Service) figure, investor, and advisor, Jason Lemkin, has kept the chat receipts and posted them on X/Twitter. Naturally, Lemkin says they won’t be trusting Replit for any further projects.

Comment Difference between this and keeping data in memory (Score 1) 36

by protopatterns (#65487462) Attached to: VP.net Promises "Cryptographically Verifiable Privacy"
What's the difference between this and expressvpn, which (supposedly) only keeps your all data --- including your IP --- in ephemeral memory? I see that as much less risky than keeping everything in 'hardware safes', even if they're encrypted: it is still a persistent form of memory. What I'm asking: is this really better? (honest question, trusting the company aside)

Comment Re:Does not compute... (Score 1) 90

they still seem to be Ad driven, so again, where is that revenue if access is purely by API?

From the (blurb to the) article: 'Stack Overflow's new deal giving OpenAI access to its API as a source of data'. Stack Overflow doesn't give away the info they have for free. Access may be purely by API, but that says nothing about how SO charges for API access when they deal with a giant company like OpenAI.

Comment LLMs can be good (Score 1) 32

by protopatterns (#64357242) Attached to: Congress Bans Staff Use of Microsoft's AI Copilot
At first glance this appears obvious: you don't want any amateur stuff in coding for cybersecurity. Most of the net is comprised of this, and LLMs are for the most part trained on the net. But a good programmer can do quite a bit with refreshing their memory via LLMs, then actually working on whatever problem they have to solve.

The major hurdle, at least as far as I see, is that institutions have to make blanket statements to prevent people (those few? hope so) from getting lazy, especially under time constraints (e.g. 'Looks good enough to me! Besides it's passed all our automated tests so I'll check it more later' and plunk! It sinks into the memory-hole).

Comment 'Even' insiders? (Score 1) 40

by protopatterns (#63919107) Attached to: Even Google Insiders Are Questioning Bard AI Chatbot's Usefulness
I wish it weren't so, but every single software company I've worked for, I would stay away from. Because I've seen the sausage getting made (!) Most of it from rotting parts: we're just there to keep the meat looking fresh, by whatever coding means. I think that LLMs do in fact have promise. Just not yet. I have a lot of reasons why neural networks aren't the whole picture, but imo they are definitely a step in the right direction. Just probably not the last step.

Submission + - A primary marker at Crawford Lake, Canada, starts the Anthropocene epoc (sciencedaily.com)

Submitted by pyroclast
pyroclast writes: The Anthropocene Working Group have put forward Crawford Lake, in Canada, as a Global Boundary Stratotype Section and Point (GSSP) for the Anthropocene. A GSSP is an internationally agreed-upon reference point to show the start of a new geological period or epoch in layers of rock that have built up through the ages.

The team has gathered core sample sections from a variety of environments around the world, from coral reefs to ice sheets. Samples from a range of these sites were then sent for analysis to the University of Southampton's GAU-Radioanalytical labs at the National Oceanography Centre in Southampton. Researchers there processed the samples to detect a key marker of human influence on the environment — the presence of plutonium.

Professor Andrew Cundy, Chair in Environmental Radiochemistry at the University of Southampton and member of the Anthropocene Working Group, explains: "The presence of plutonium gives us a stark indicator of when humanity became such a dominant force that it could leave a unique global 'fingerprint' on our planet.

Other geological indicators of human activity include high levels of ash from coal-fired power stations, high concentrations of heavy metals, such as lead, and the presence of plastic fibres and fragments. These coincide with 'The Great Acceleration' — a dramatic surge across a range of human activity, from transportation to energy use, starting in the mid-20th century and continuing today.

Comment Re: lol (Score 1) 293

Sure, no system is perfect but I would submit a few things as facts:

1. Canada's healthcare on teh whole is more equitable and affordable than the US system. 2. Canada's system is on better footing than the US and can way more easily adapt it's system to the rising costs. 3. Canada's populace would never swap out to the US system but would probably move to a system more like the Nordics/Dutch/Germany which are all better.

I am not going to defend every aspect of Canada's healthcare but I will not stand for people using it to defend the US system which is my opinion indefensible.

Yes (..no system is perfect, but this one's a better option. I know because I've been through Canada's health system)

Submission + - Flipper hacking device on track to make $80M worth of sales (techcrunch.com)

Submitted by ZipNada
ZipNada writes: Flipper Zero is a “portable gamified multi-tool” aimed at everyone with an interest in cybersecurity, whether as a penetration tester, curious nerd or student — or with more nefarious purposes. The tool includes a bunch of ways to manipulate the world around you, including wireless devices (think garage openers), RFID card systems, remote keyless systems, key fobs, entry to barriers, etc. Basically, you can program it to emulate a bunch of different lock systems.

The system really works, too — I’m not much of a hacker, but I’ve been able to open garages, activate elevators and open other locking systems that should be way beyond my hacking skill level.

Slashdot Top Deals

System going down in 5 minutes.

Close