Fiat currencies and central banks have been the biggest boon to the employment of Keynesians. Carney's faith in "managing" the economy with a more nuanced leviathan is only exceeded by his exuberance in ignoring the relationship of high density energy adoption (as opposed to replace it with lower density energy) with prosperity and climate resilience.
“We will not have more crashes in our time.” - John Maynard Keynes, 1927
"There was one good thing about Marx: he was not a Keynesian." - Murray Rothbard
Thank you for the information. Seems a generous assessment, from my perspective. I would have liked to have seen an impartial review for Carney’s decision to sell off all of Canada’s gold reserve. I would also have appreciated a breakdown of some of the projects undertaken by Brookfield while he was CEO, for example clearing of Brazil’s rainforest or investment in pipelines. I found some of the policies he outlined in his book to be a cut and paste from the manifesto of the WEF. If positing a position regarding what sort of Prime Minister Mr Carney would make, it would also be pertinent to include an analysis of how a person can hold multiple allegiances and act solely for the good of the country he means to govern. At the end of the day the AI model seems to ignore any conflict of interest that supersedes its own analysis
I hear you, but I question whether most people know as much about Poilievre as they think they do. He is good at spin (as a long time politician) so it can be hard to understand what he is really saying (and people tend to hear what they want to hear). If Poilievre has a shorter paper trail, the exercise should be quicker! Anyway, I look forward to reading it, if you decide to do it.
I've actually got a lot of experience with LLMs (I wrote a book on generative AI), and I have sometimes seen left bias. But these specific Carney-related responses were quite candid about some of the serious vulnerabilities of Carney's past positions and how they contributed to bad decision-making.
Also, I have to admit that I really enjoy talking with AI chat tools about politically-charged topics. In my experience they've always been fully responsive to counter arguments and getting them to admit their initial assumptions might not be correct is satisfying. :)
Sounds good. Since you know the nuts and bolts of how to work these things under the hood, what specific measures have you used to check for and correct for political bias in your LLMs?
Well unless we're able to build our own LLM from scratch - which is a bit out of my comfort range even with open source access to the new DeepSeek model - there's no way to "fix" bias. It's *possible* that even OpenAI engineers - assuming they wanted to - couldn't do a perfect job there. But I've found that a moderate amount of domain knowledge and experience gives you the tools to push back with clarifying followup questions.
Also, making the initial prompt as clear as possible, with whole lists of conditions and counterfactuals ("taking x into account and allowing for the possibility of y...") usually improves the quality of the responses.
I might (once we get closer to an election). But there are two caveats:
1. The country already knows a lot about Pierre Poilievre and his plans for governance because of his many years in Parliament - so there's (probably) less mystery.
2. Thanks to Carney's academic background, he's probably left behind a longer *paper trail* of policy thinking than Poilievre - so there's not as much about Poilievre to research.
I would like to know if Covid lockdowns played a roll in decreasing the liklihood of the health system becoming overwhelmed and possibly collapsing. Could your AI app be used to explore that question?
Well this risks sending me (back) down a VERY deep rabbit hole. But here's the link to a chat I just had comparing the experiences of Sweden (which had no lockdown and suffered lower all-cause mortality rates over the COVID years than peer countries) and peer countries: https://chatgpt.com/share/6798e5a3-8770-8001-b609-06ce40be8a69
Fiat currencies and central banks have been the biggest boon to the employment of Keynesians. Carney's faith in "managing" the economy with a more nuanced leviathan is only exceeded by his exuberance in ignoring the relationship of high density energy adoption (as opposed to replace it with lower density energy) with prosperity and climate resilience.
“We will not have more crashes in our time.” - John Maynard Keynes, 1927
"There was one good thing about Marx: he was not a Keynesian." - Murray Rothbard
Thank you for the information. Seems a generous assessment, from my perspective. I would have liked to have seen an impartial review for Carney’s decision to sell off all of Canada’s gold reserve. I would also have appreciated a breakdown of some of the projects undertaken by Brookfield while he was CEO, for example clearing of Brazil’s rainforest or investment in pipelines. I found some of the policies he outlined in his book to be a cut and paste from the manifesto of the WEF. If positing a position regarding what sort of Prime Minister Mr Carney would make, it would also be pertinent to include an analysis of how a person can hold multiple allegiances and act solely for the good of the country he means to govern. At the end of the day the AI model seems to ignore any conflict of interest that supersedes its own analysis
I hear you, but I question whether most people know as much about Poilievre as they think they do. He is good at spin (as a long time politician) so it can be hard to understand what he is really saying (and people tend to hear what they want to hear). If Poilievre has a shorter paper trail, the exercise should be quicker! Anyway, I look forward to reading it, if you decide to do it.
It's an interesting project and I appreciate your work. You might need to take AI output with a hefty grain of salt though when it comes to evaluating politicians and policies. Studies tend to find that almost all LLMs exhibit a strong left-leaning bias. For example https://manhattan.institute/article/measuring-political-preferences-in-ai-systems-an-integrative-approach
I've actually got a lot of experience with LLMs (I wrote a book on generative AI), and I have sometimes seen left bias. But these specific Carney-related responses were quite candid about some of the serious vulnerabilities of Carney's past positions and how they contributed to bad decision-making.
Also, I have to admit that I really enjoy talking with AI chat tools about politically-charged topics. In my experience they've always been fully responsive to counter arguments and getting them to admit their initial assumptions might not be correct is satisfying. :)
Sounds good. Since you know the nuts and bolts of how to work these things under the hood, what specific measures have you used to check for and correct for political bias in your LLMs?
Well unless we're able to build our own LLM from scratch - which is a bit out of my comfort range even with open source access to the new DeepSeek model - there's no way to "fix" bias. It's *possible* that even OpenAI engineers - assuming they wanted to - couldn't do a perfect job there. But I've found that a moderate amount of domain knowledge and experience gives you the tools to push back with clarifying followup questions.
Also, making the initial prompt as clear as possible, with whole lists of conditions and counterfactuals ("taking x into account and allowing for the possibility of y...") usually improves the quality of the responses.
Hi David, This backgrounder is really helpful, even if not perfect. Will you do the same for Pierre Poilievre?
I might (once we get closer to an election). But there are two caveats:
1. The country already knows a lot about Pierre Poilievre and his plans for governance because of his many years in Parliament - so there's (probably) less mystery.
2. Thanks to Carney's academic background, he's probably left behind a longer *paper trail* of policy thinking than Poilievre - so there's not as much about Poilievre to research.
I would like to know if Covid lockdowns played a roll in decreasing the liklihood of the health system becoming overwhelmed and possibly collapsing. Could your AI app be used to explore that question?
Well this risks sending me (back) down a VERY deep rabbit hole. But here's the link to a chat I just had comparing the experiences of Sweden (which had no lockdown and suffered lower all-cause mortality rates over the COVID years than peer countries) and peer countries: https://chatgpt.com/share/6798e5a3-8770-8001-b609-06ce40be8a69
Thanks for the info.