We can choose to expand copyright to make #ChatGPT et al pay. (I have strong reservations about unintended consequences, but we can try.)
Or we can explicitly say you may give your AI anything you like to read, scrape away. As long as it is generating rather than repeating, a-okay, let a million LLM flowers bloom, bring on the Singularity.
The first path makes for more bespoke, hand-crafted writing by those of us who have slaved for many years in the language mines, and so deserve our oligopoly. Do not fear, citizens, we shall keep you safe from bad signal-to-noise ratios, just pay your protection money and leave it with us.
The other, well, depends on how quickly those #LLMs get good, innit, and to what extent we can disincentivize stuff like spam e-books and bullshit websites.
My email volume must be up several million percent since the late 1990s. The spam I see is down in absolute terms.
Ditto web browsing. I have been known to rail mildly against Google's take of what constitutes "quality content", but it sure down-ranks a lot of awful stuff.
I realise Twitter has failed, but that's why we're here, right?
@phillipdewet He's great. And Line Goes Up is peak. Once you're done with that, go see In Search of a Flat Earth. It's interesting to see his skills develop from the beginning of his channel.
@phillipdewet part of this is how to get the unheard voices to the table - ie more writers, beyond the current group. One of the criticisms of the current LLM state is the bias in the training material, which is a consequence of who gets to put up/ keep/take down things like Wikipedia articles (W being a major training source.)
@grant_h I increasingly think LLMs are the best way to get more people included.
A first draft is hard. Given a starting point, though, many people can make it their own.
And it doesn't matter how bad – or how biased – that first draft is. It is hard to teach people to write. You know what is easy? A contrary reading. A 3-year-old can do a contrary reading. Everyone is a frigging contrary reader.
It's the social-media-engagement hack: give 'em a wrong answer and they'll write a correct one.
@phillipdewet this is true. It lowers the barrier to entry, but we still need to work on making entry points for people. Financial viability is another parameter in a set of many.
@phillipdewet Writing of any sort - fiction, news, research, takes time. Time they could have been earning money to pay the bond/...
There are financially secure people who can opine, but part of creating an ecosystem where more voices are heard requires that there is at least some money in it, to include people who aren't so secure. A "reasonable" reward for the effort is required.
It would be alongside a number of other parameters - editing support, encouragement, security for reporting 1/n
"Reasonable" is interesting here. Reward systems seem skewed to getting the already financially secure to write more; sufficiently large sums of money to lure them out of their opium dens and heated swimming pools and whatnot.
Retuning for people with more moderate needs would cause an uproar – and could be a very good thing.
Add comment