Fair Training, Not Forced Costs – Stop the SIA Refresher

Dear Bill, what do you think about stealing now? I was rereading John Markoff's excellent and illuminating What The Dormouse Said, about the role psychedelics and Sixties counterculture had in shaping the early personal computing industry and came across this open letter from Bill (Then at Micro-Soft with a hyphen!) in the Homebrew Computer Club newsletter. He lays out his frustrations with the unauthorised free (and sometimes paid) distribution of the Altair BASIC software he and Paul Allen and Monte Davidoff had created, something he calls out as theft. It's a compelling argument and one that feels very relevant today reading the excellent work being done on #IPtheft and #AI by the likes of Johan Cedmar-Brandstedt, Virginie Berger, Evan Winston, Ann Handley, Ed Newton-Rex and my CSC Global Alumni colleague Dr Elizabeth Mkoba. As Gates notes in his letter, without fair payment for their work, there won't be great software in the future, in the same way that by not paying artists you use for training we will have less great art both to enjoy but also to train on and by depriving websites of traffic with AI summaries you will end up with less great content both for us to read but also to summarise. Stealing doesn't build sustainable knowledge economies. It seems to work if you're colonising somewhere and extracting resources (See history for more details) but it's obviously morally abhorrent and arguably now being shown as unsustainable in the long run as well. And it won't work here. That's why we have intellectual property law as the knowledge economy equivalent of property law (And as others have pointed out, it's unlikely that Jack and Elon would enjoy it if we deleted all property law). The argument against it is that development would have been slower but when you look at the pace AI development has been moving (And the money that has been spent) a few months or even a couple years more (At worst) would not make a material difference to humanity or governments and economies, it really only matters to stock prices. Constraint breeds innovation. DeepSeek have still used stolen training data but they didn't have access to the resources that conventional wisdom said were absolutely necessary to build an LLM and look where we are. So let's not build the future on the same morally abhorrent foundations we built the past on, because it may seem quicker and easier at the time but it will come back to bite you at some point. I'd therefore love to see Bill take a similar stand to this again and back a fair application of copyright and IP law. Full text of the letter (Too long for Alt-text) is in the first comment. I hope this constitutes fair use otherwise: #irony. #FairTraining | 27 comments on LinkedIn