Two giant strides in the artificial intelligence industry— one achieved, another threatened—a rekindling existential anxieties. The first is the debut of ‘Tilly Norwood’, an AI-generated ‘actress’. Actors have strong feelings about this, but her birth is not really a watershed. The other development is the latest skirmish in Elon Musk’s war with Wikipedia, which is going on doggedly.
Musk threatens to launch the awfully named ‘Grokipedia’, powered by his Grok AI, to take on the site that had audaciously—and mostly successfully— tried to become the crowdsourced repository of all of human knowledge. This is a serious development because it proposes to put machines in charge of ‘correcting’ human knowledge and ironing out bias, which is an exclusively human concept. The universe doesn’t know about bias. It doesn’t tolerate it, either.
In comparison, Tilly Norwood is old hat. For ages, entertainment has been yearning for technology to extend the life of franchises beyond the lifetime of their actors. Automatons were test-flighted long ago. In 2000, the Press Association of the UK debuted the virtual newsreader Ananova, who built up an impressive following. Visually, Ananova was only as sophisticated as characters of contemporary video games like Grand Theft Auto. In contrast, Norwood has every hair in place. She also has her own social media profiles. From the general public’s vantage, she is exactly like celebrities—elevated beings we see only on screen and interact with only on social media. Naturally, talent scouts are interested.
Should this bother actors? Indeed, professionals who have not quite become brands would lose work to Norwoods. But in formula entertainment, the bread and butter of the trade, acting skills are optional. Clothes-horses with basic facial expressions drive some of the most lucrative serials. Privately, directors and producers have always yearned to replace difficult actors with automatons bereft of tantrums, who are available every day on time. Tilly was an engaging idea decades before AI. People have had time to get used to it.
On the contrary, Grokipedia represents paradigm change. The latest skirmish in Musk’s feud began with conservative host Tucker Carlson talking with the knowledge site’s cofounder Larry Sanger, who shaped Wikipedia’s editorial philosophy and the revolutionary wiki-based community structure. Sanger revealed Wikipedia’s blacklisted sources—Breitbart News, The Daily Caller, The Epoch Times, Fox News, The New York Post, The Federalist—and a whitelist of ‘perennial sources’—The New York Times, The Washington Post, CNN, The Nation, Mother Jones, GLAAD. This apparently rekindled Musk’s long-term interest in ‘far-left’ Wikipedia’s ‘bias’. Now, with Grokipedia, which was trained on Wikipedia data among other inputs, he plans to publish a cleaned-up version of Wikipedia. Wouldn’t it make this ouroboros-like project biased, too?
Perceptions of bias depend on where you stand. If your point of view is far to the right, even an object near the centre of the spectrum could look far left. Plus, an entity’s choices and associations, like Wikipedia trusting certain media and not others, does not signal bias. It’s ideology or politics. Neither is a bad word.
It is normal for the politics of groups defined by history, culture, interests, ideals, and imagined futures to differ. Democracy is built upon negotiations between these very differences. When the ruling party and the opposition debate in parliament, they negotiate a common minimum programme for running the country. Its elements rest on a sliding scale stretching between the extremes of what the ruling party thinks is ideal, and what the opposition sees as totally unacceptable. The scale is measured out by discussion, not by assigning an external authority, like an omniscient machine, to decide what is good or bad according to the ideology of the ruling party.
If that actually worked, not only could AI clean up Wikipedia, it could also run teashops, corporations, and even nations. But it can’t, because right and wrong, good and evil, nice and awful are human concepts, which only humans can evaluate, however imperfectly. And there is nothing outside of the empirical sciences of the gross world—like the specific heat of water or the atomic number of plutonium—which is exactly right or wrong.
Phenomenons outside the STEM world, like stock markets and politics, run on sentiments and narratives, not absolute facts. This is one of the reasons why, in the politics of so many nations, the good guys who respect facts are now on the back foot, mercilessly baited by liars and fabulists, and unable to regain the psychological advantage. An ideology can be pure fiction and still be consistent, and serve as a myth that determines current affairs.
Wikipedia was created to democratise knowledge and enable informed choice. A Grokipedia would not address its deficits because, in a polarised world, ‘alternative facts’ are also facts. There’s nothing to clean up in the domain of facts, but it is very easy to influence worldviews. After that’s done, perhaps a journalist like Ananova will report the new reality, and the engaging Tilly will play the lead in the serial that follows the biopic, video game, or election campaign in the making. Perhaps we are already led by flesh-and-blood Tilly Norwoods who are preparing us for a worse future.
Pratik Kanjilal | SPEAKEASY | Senior Fellow, Henry J Leir Institute of Migration and Human Security, Fletcher School, Tufts University
(Views are personal)
(Tweets @pratik_k)
