People with too strong statements of future AI capabilities are just not worth listening to. Why? Because we just don't know. To claim architects won't be replaced is might very well just be a dream
@sandman89176 yes, and I was also including LLM which have shown new, unexpected and remarkable features as the models have grown. We have a hard time predicting what new model sizes will bring and thus how different jobs will e affected
@@dahla1973model sizes have the problem of data quality and it seems (can't recall the article, sorry) that biggest LLM companies already reached a sort of plateau. It'll be interesting to see how they adapt and I agree with you when you say that currently we have no means to predict precisely future evolutions
@dahla1973 - my understanding of Avrahams "strong statements" were that he was not doubting future capabilities of AI - he was pointing out that Humans have a critical role in decision making - the keywords he used where morals, responsibility and accountability. AI can be used as an input to the analysis and decision making process to find issues and trade-offs and resolve them, it can be part of the feedback loop as we finalise what we want the system to do and how it should do it - but it should NEVER be the decision maker - humans are responsible to ensure the system is safe and useful for humans. I'm tempted to add that people who disagree with that are just not worth listening to - but that would be rude so I'll refrain from doing that ;-)
Great talk Avraham!
People with too strong statements of future AI capabilities are just not worth listening to. Why? Because we just don't know. To claim architects won't be replaced is might very well just be a dream
AI in this talk was specifically referring to LLMs
@sandman89176 yes, and I was also including LLM which have shown new, unexpected and remarkable features as the models have grown. We have a hard time predicting what new model sizes will bring and thus how different jobs will e affected
@@dahla1973model sizes have the problem of data quality and it seems (can't recall the article, sorry) that biggest LLM companies already reached a sort of plateau. It'll be interesting to see how they adapt and I agree with you when you say that currently we have no means to predict precisely future evolutions
@@Khaervek yes, llms might have hit a plateu. However unclear at this point why and if it can be overcome. Might just be a temporary issue
@dahla1973 - my understanding of Avrahams "strong statements" were that he was not doubting future capabilities of AI - he was pointing out that Humans have a critical role in decision making - the keywords he used where morals, responsibility and accountability. AI can be used as an input to the analysis and decision making process to find issues and trade-offs and resolve them, it can be part of the feedback loop as we finalise what we want the system to do and how it should do it - but it should NEVER be the decision maker - humans are responsible to ensure the system is safe and useful for humans. I'm tempted to add that people who disagree with that are just not worth listening to - but that would be rude so I'll refrain from doing that ;-)