anaguerberofFeb 9, 2023
Jelle Zuidema

How do we wake up regulators about the risks of the upcoming 'conversational' search engine war?

I have, as many in NLP worldwide in the last 2 months, talked to many journalists and academics from other fields about #ChatGPT. Many of us have warned rigth away (e.g., [1]) that the commercial interests & competition in big tech will lead to this technology being unrolled prematurely, without proper independent risk assessment. Is anyone doing anything about it?

[1] https://www.volkskrant.nl/nieuws-achtergrond/een-moordscene-a-la-nicci-french-gemaakt-met-kunstmatige-intelligentie-en-meer-iedereen-is-verbijsterd-over-de-kwaliteit~b6e6df4e/

Een moordscène à la Nicci French gemaakt met kunstmatige intelligentie, én meer: ‘Iedereen is verbijsterd over de kwaliteit’

Een doortimmerd marketingplan maken, een column voor de krant schrijven of een bestaande computercode verbeteren: het taalprogramma ChatGPT verbaas...

de Volkskrant
Feb 8, 2023 at 11:07amWeb
Show threadJelle ZuidemaFeb 8, 2023
I worry, however, that many don't take those risks very seriously: a little bit of plagiarism, some misinformation, a bit of misogyny -- nothing new, right? That's a mistake when we talk about search engines: they are the entry point to the internet. When Facebook, another key entry point, changed its policies several times in the last decade, this had massive direct (e.g., on internet traffic to newspapers) & indirect effects. When Bing/Perplexity/You & others now integrate search with
Show threadJelle ZuidemaFeb 8, 2023
#chatgpt-like techniques & force Google to the same, we don't know the consequences, but accidents are bound to happen: companies go bankrupt, crucial information does not reach key people, while misinformation does. It's as if a company controlling a major highway suddenly redirects all traffic to a secret road; surely, governments would want a say? I'm no expert on regulations but I wish regulators would hit a pause button & create an opportunity for some proper auditing of this technology.
Show threadJelle ZuidemaFeb 12, 2023

I've sometimes disagreed with @garymarcus's wide ranging critique of deep learning, but this piece is again really good and a powerful warning. It ends with a wake-up call for regulators (as my thread above): "And, bonus, there is little if any government regulation in place to do much about this. The possibilities are now endless for propaganda, troll farms, and rings of fake websites that degrade trust across the internet.".

https://garymarcus.substack.com/p/inside-the-heart-of-chatgpts-darkness

Inside the Heart of ChatGPT’s Darkness

Nightmare on LLM Street

Marcus on AI
Show threadErik JonkerFeb 8, 2023
@wzuidema totally agree, if somebody wants to hit that "pause button", at least for EU for starters, quick action is necessary, all the risks you mention are probably the reason Google is moving slower (a lot to lose) then Microsoft (nothing to lose with regard to search)
Show threadJelle ZuidemaFeb 8, 2023

@ErikJonker Yes, curious whether the EU indeed can act & dares to (it might not make them more popular with European citizens in the short term).

Fortunately, Microsoft does have a lot to loose w.r.t. reputation for all of its other products. Reputation is the main driver keeping them at leats somewhat in check, I reckon.

Show threadErik JonkerFeb 8, 2023
@wzuidema true, especially if you think about the Azure/cloud market, dutch government is a big customer, they have something to loose there