Post #2198883

@libramoon@mastodon.social @tynstar@nerdculture.de @dx@social.ridetrans.it There appears to actually exist several schools of thought on these question among the AI industry. Let’s take 3 examples: 1. Yann Le Cun calls it full BS with good reasons: lack of world representation, current limitations to text input for both interaction and training etc. However he sells his own "world model" and tries to raise money to make even more capable AI. 2. Dario Amodei basically builds all his business arguments around the fact that if AI research continue, AGI is inevitable EXCEPT if you fund HIS AI research so that HE can develop safeguards against it. This is, IMHO, the most stupid argument of all I've heard, but it works on the investors and the public. 3. Yoshua Bengio, which is a leading expert in AI security, warns that AI agents can definetely act against instructions but is never warning against AGI. More about people relying on defective tools on the one hand and malignant use from actors trying to influence people opinion. 4. Sam Altman is pushing forward for AGI but with obvious financial incentive given that his business model is still a bug financial black hole on which a substantial portion of the US economy rely. So meh. Is AGI really a risk? Who cares. Are experts warning against it? Hard to see a consensus. Are some actors in the field using it as a red scarf to get people to put money in there shit? Definetely.