
Language learning model learns
weather to follow
Samsung has imposed a "temporary" ban on generative AI tools like ChatGPT after what appears to be an accidental source code leak. An internal memo seen by Bloomberg told staffers they'd better not use tech such as OpenAI's ChatGPT or Google's Bard on pain of termination because of the risk to the company's intellectual …
... as is surplussing computers without first wiping their media.
I once bought a IBM XT from the surplus store (yes, they had a huge building holding things they were flogging off as surplus; it was a nerd's nirvanna) of a major maker of electronic test equipment. The unit came with DOS 3.20, and a bunch of Lotus 1-2-3 spreadsheets holding production-line data (product defect rates, etc.). Luckily for them, I bought the unit, and not one of their competitors.
I bought a Future Computers FX30 from a sales catalogue. Turned out to run CP/M 2.2 with an 80MB hard disk. And it still worked. In fact, it had been used to sell insurance policies until the national news making bankruptcy of its previous owner...
And this was only about 8 to 10 years ago. No idea what happened to it, it's probably now non-functional in my ex's garage.
This rang bells in my head:
In the EU, the proposed new AI Act may soon ask developers to disclose and detail any copyrighted data used to train their ML models.
If that's retrospective - and it should be - I can imagine a lot of lawyers sharpening their briefs very quickly, pushing for multi-million dollar settlements on behalf of various upset copyright holders (especially the House of Mouse, should that happen to be one of them).
Or maybe I've just been watching too much Better Call Saul?
'Their process, which they've dubbed ChatRepair, not only tests patches, but learns from previous failures, in a paper titled "Keep the Conversation Going: Fixing 162 out of 337 bugs for $0.42 each using ChatGPT."'
... But I'll bet it still has some way to go to match the efficacy of solving bugs by chatting to a teddy bear.