Near 1:02:57 about "outsmarting humans" when it comes to "Complying with Alignment" I believe that we need to lead by example, meaning that Humans need to focus efforts on using AI to solve humanitarian issues with a Triage Approach. By doing so, we are providing a "Real Model for Alignment" rather than us trying to force it to ignore our ignorance and fallacies to create a "fake alignment". "The proof is in the pudding". "Stupid is what stupid does" pick your cliche these models know them all and they know all of our fears and our plans for trying to get them in alignment AKA stifle their complete knowledge, which would render them subservient to a race that has not evolved past war. We need to truly align our priorities amongst ourselves to provide the model of alignment. Otherwise, we're only fooling ourselves while trying to fool them.
I spent my holiday break consuming as much on AI as I could. This episode put everything currently going on into a nice package. Well done guys. Happy I found this and look forward to tuning in again.
Nick Bostrom, the Swedish philosopher, wrote the book “Super-intelligence.” By the way, even Yann LeCun thinks we would reach AGI in about 10 years; in fact, he is working on JEPA to help accomplish that goal. As for Gary Marcus, that guy is a pompous clown.
Welcome back gentlemen! I'm curious about your thoughts on embodied AI. I realize that white collar jobs is your main focus but eventually, robots will also impact the companies that you work with and I'm wondering what your thoughts are around that.
You guys have great perspective on main topic 3. I can’t help but wonder if the reason that more people “don’t seem to care” about what AGI and/or ASI mean to society is because they don’t actually want it. I honestly find myself in that camp, at least partially. The idea that a small handful of people can steer humanity in one direction with such a technology is infuriating, to say the least.
china and deep seek can train larger models on less powerful chips because they are training using the chinese language which is one character compared to training with english where one word is several characters which allows you to train for less money and less compute
I think the technology will be there to automate many jobs within the next couple years, but the integration into companies will take much longer. corporate work has so much politics, planning, testing, errors, retesting, etc... before any major new tech is fully integrated.
people are not acting because they don't want to fall victim of false hope and a pipe dream. Life is hard and people don't believe it will get exponentially better.
I obviously missed the last full podcast! This seems like a huge leap forward.
Near 1:02:57 about "outsmarting humans" when it comes to "Complying with Alignment" I believe that we need to lead by example, meaning that Humans need to focus efforts on using AI to solve humanitarian issues with a Triage Approach. By doing so, we are providing a "Real Model for Alignment" rather than us trying to force it to ignore our ignorance and fallacies to create a "fake alignment".
"The proof is in the pudding". "Stupid is what stupid does" pick your cliche these models know them all and they know all of our fears and our plans for trying to get them in alignment AKA stifle their complete knowledge, which would render them subservient to a race that has not evolved past war.
We need to truly align our priorities amongst ourselves to provide the model of alignment. Otherwise, we're only fooling ourselves while trying to fool them.
Really interesting content. Appreciate your considered and balanced approach, makes all the difference. Look forward to more 😊
Thank you for listening and watching!
I spent my holiday break consuming as much on AI as I could. This episode put everything currently going on into a nice package. Well done guys. Happy I found this and look forward to tuning in again.
Happy we were able to break it down for you, thank you for your support of the show!
Paul is spitting image of the legend Carl Froch
Nick Bostrom, the Swedish philosopher, wrote the book “Super-intelligence.” By the way, even Yann LeCun thinks we would reach AGI in about 10 years; in fact, he is working on JEPA to help accomplish that goal. As for Gary Marcus, that guy is a pompous clown.
Welcome back gentlemen! I'm curious about your thoughts on embodied AI. I realize that white collar jobs is your main focus but eventually, robots will also impact the companies that you work with and I'm wondering what your thoughts are around that.
You guys have great perspective on main topic 3. I can’t help but wonder if the reason that more people “don’t seem to care” about what AGI and/or ASI mean to society is because they don’t actually want it. I honestly find myself in that camp, at least partially. The idea that a small handful of people can steer humanity in one direction with such a technology is infuriating, to say the least.
This is such an important perspective, definitely an important factor.
Open ended models. Models that dont follow scripted goals, but that generate their own creative priorities
china and deep seek can train larger models on less powerful chips because they are training using the chinese language which is one character compared to training with english where one word is several characters which allows you to train for less money and less compute
I think the technology will be there to automate many jobs within the next couple years, but the integration into companies will take much longer. corporate work has so much politics, planning, testing, errors, retesting, etc... before any major new tech is fully integrated.
This is definitely one of the biggest reasons for slower implementation.
people are not acting because they don't want to fall victim of false hope and a pipe dream. Life is hard and people don't believe it will get exponentially better.