This is not only a demonstration of AI's writing ability, but also a mirror that reflects the dilemma of our political choices. ChatGPT's analysis really captures the essential differences between the two presidents' governing styles and policy priorities: one emphasizes gestures and breaking the old and establishing the new, while the other is committed to restoring normalcy but is led by progressives. Amazingly, this "editorial" has almost no obvious emotions, but it is full of judgments, like a sophisticated editor holding a pen and weighing the weight of each word. This also raises a deeper question: when AI can simulate media discourse so accurately, do we also have to redefine "objectivity" and "balance", the cornerstones of journalism? I especially agree with the open question you raised at the end: not whether AI can write editorials, but whether it should write? At a time when true and false information is mixed and human bias is inevitable, perhaps AI is not a threat, but another possibility - provided that we still have the freedom to judge and question. Thank you for using this experiment to evoke a meditation on the media, fairness and the future
I don't have the hard data but I would tend to believe that it would be hard to get a politically balanced, non-partisan *psychological* profile as was asked of the AI. The issue is with the training data and as soon as you invoked psychology the pendulum was already moved left. AI could do the best it could to return to center but I believe that the bias would always be left leaning due to the training data the AI was built on. As a case in point, I recently attended the psychology class graduation for MSU. I was expecting a left lean but was astounded by how bad it really was. Within the first 1-2 sentences of the commencement the very left lean of the school was in plain view. In fact, it was clear there zero genuine opportunity for open dialog or dissenting thoughts from the graduates or even the audience who were even queried for their opinion (agreement) at some points. It was so bad that I looked up MSU's rules for these speakers and found that the most significant rules designed to keep these events non-partisan were openly broken. I found this very disappointing but maybe it should have been expected from higher learning in America today? Based on this (limited) experience plus the likely left lean of published content from this field, my belief is that the AI the first time around would have been struggling to find center.
This second attempt is far more balanced... perhaps too balanced for some but I think overall does a better job of raising the pros and cons of both men and their administrations. It is certainly more in line with traditional journalism that provides the readers with the facts and then lets them come to their own conclusion.
Right at the start, the speaker (I believe the dean) made a comment along the lines of "Look at all of these wonderful... woke... graduates" with an intentional emphasis on the word "woke". It felt like a loaded term being used in a directed way, as if to imply a shared ideology was required among the graduates without directly stating it. This struck me as an odd and unnecessarily partisan way to begin what should have been a unifying event.
Later, a former professor took the stage and spent several minutes on what I can only describe as a political diatribe against the United States. He painted the country in a very negative light, making multiple partisan points, and even asked for (requiring) audience affirmation of his beliefs. From my understanding, this is a violation of MSU’s own commencement guidelines which require one-way speeches that do not solicit responses from the audience.
Overall, I was very disappointed in MSU and that the event couldn’t remain focused on what it was meant to celebrate... the accomplishments and bright futures of the graduates. Thankfully, the student speakers were a highlight as they kept their remarks thoughtful, hopeful, and largely non-political. That gave me some optimism that maybe the next generation understands the value of unity in moments that are meant to bring people together!
Right or wrong, a lot of institutions consider themselves under attack. I was at the American Library Association national convention last year and there was outright anger in the crowd of 2,000+ about the book bannings in schools. But a college graduation is no place to invoke political issues--especially asking the audience for a show of support.
Very interesting. What I see in ChatGPTs analysis is a bias against messy but necessary change. It seems to say “Biden was less controversial but largely ineffective”. Conversely it implies “Trump tried to address some major issues, but failed to achieve unity.”
If one prompted ChatGPT re the premise that America’s successful future will rely on change in our economic structure, Trump policies upsetting institutions might be positioned as a necessary cost.
Conversely, if ChatGPT were prompted to focus on achieving more equal outcomes across different population minorities (including the white males), Biden woukd likely be viewed as more successful.
So IMO it’s the objective given in the AI prompt that matters the most. Which do Americans want more a) economic prosperity that lifts some more than others or b) equality of outcomes that tries to eliminate winners vs losers?
It seems to have some moderate conservative bias and it completely missed the causes of inflation which was over spending on stimulus spending which was caused by Trump not Biden.
Basically December 2020 and January 2021 saw huge amounts of stimulus happen causing the situation where you have too much money chasing too few goods. No one saw the problems at the harbor which also caused normal shipping to slow to a crawl. It did not talk about the fact that an additional 400,000 Americans died because of Trump's pandemic response.
Finally the best and shortest synopsis of the difference between Trump and Biden can be summed up by the Economists endorsement of Kamala Harris on 10/31/24.
"While some newspapers refused to back a presidential candidate this year, today The Economist is endorsing Kamala Harris. Tens of millions of Americans will vote for Mr Trump next week. Some will be true believers. But many will take a calculated risk that in office his worst instincts would be constrained. We see that as recklessly complacent. By making Mr Trump leader of the free world, Americans would be gambling with the economy, the rule of law and international peace. Ms Harris’s shortcomings, by contrast, are ordinary. And none of them are disqualifying. If The Economist had a vote, we would cast it for her."
This was my first poll, so I'm not sure why it's not working for you. I'm guessing only subscribers can take the poll--but not sure about this. I'll investigate.
I am impressed by CHATGPT every day. Clearly the training data will influence the results, and not having access to that leaves one wondering. Thus it seems the answer can be predicted by the control of inputs as training data, so how does one make that unbiased and complete ? We all have friends that watch FOX all day, and we know how those conversations go.
This is not only a demonstration of AI's writing ability, but also a mirror that reflects the dilemma of our political choices. ChatGPT's analysis really captures the essential differences between the two presidents' governing styles and policy priorities: one emphasizes gestures and breaking the old and establishing the new, while the other is committed to restoring normalcy but is led by progressives. Amazingly, this "editorial" has almost no obvious emotions, but it is full of judgments, like a sophisticated editor holding a pen and weighing the weight of each word. This also raises a deeper question: when AI can simulate media discourse so accurately, do we also have to redefine "objectivity" and "balance", the cornerstones of journalism? I especially agree with the open question you raised at the end: not whether AI can write editorials, but whether it should write? At a time when true and false information is mixed and human bias is inevitable, perhaps AI is not a threat, but another possibility - provided that we still have the freedom to judge and question. Thank you for using this experiment to evoke a meditation on the media, fairness and the future
I don't have the hard data but I would tend to believe that it would be hard to get a politically balanced, non-partisan *psychological* profile as was asked of the AI. The issue is with the training data and as soon as you invoked psychology the pendulum was already moved left. AI could do the best it could to return to center but I believe that the bias would always be left leaning due to the training data the AI was built on. As a case in point, I recently attended the psychology class graduation for MSU. I was expecting a left lean but was astounded by how bad it really was. Within the first 1-2 sentences of the commencement the very left lean of the school was in plain view. In fact, it was clear there zero genuine opportunity for open dialog or dissenting thoughts from the graduates or even the audience who were even queried for their opinion (agreement) at some points. It was so bad that I looked up MSU's rules for these speakers and found that the most significant rules designed to keep these events non-partisan were openly broken. I found this very disappointing but maybe it should have been expected from higher learning in America today? Based on this (limited) experience plus the likely left lean of published content from this field, my belief is that the AI the first time around would have been struggling to find center.
This second attempt is far more balanced... perhaps too balanced for some but I think overall does a better job of raising the pros and cons of both men and their administrations. It is certainly more in line with traditional journalism that provides the readers with the facts and then lets them come to their own conclusion.
Ken, very interesting regarding the MSU commencement. What specific comments pushed the commencement to the far left?
Right at the start, the speaker (I believe the dean) made a comment along the lines of "Look at all of these wonderful... woke... graduates" with an intentional emphasis on the word "woke". It felt like a loaded term being used in a directed way, as if to imply a shared ideology was required among the graduates without directly stating it. This struck me as an odd and unnecessarily partisan way to begin what should have been a unifying event.
Later, a former professor took the stage and spent several minutes on what I can only describe as a political diatribe against the United States. He painted the country in a very negative light, making multiple partisan points, and even asked for (requiring) audience affirmation of his beliefs. From my understanding, this is a violation of MSU’s own commencement guidelines which require one-way speeches that do not solicit responses from the audience.
Overall, I was very disappointed in MSU and that the event couldn’t remain focused on what it was meant to celebrate... the accomplishments and bright futures of the graduates. Thankfully, the student speakers were a highlight as they kept their remarks thoughtful, hopeful, and largely non-political. That gave me some optimism that maybe the next generation understands the value of unity in moments that are meant to bring people together!
Right or wrong, a lot of institutions consider themselves under attack. I was at the American Library Association national convention last year and there was outright anger in the crowd of 2,000+ about the book bannings in schools. But a college graduation is no place to invoke political issues--especially asking the audience for a show of support.
Good question. I'll do that later this year after the dust settles.
Very interesting. What I see in ChatGPTs analysis is a bias against messy but necessary change. It seems to say “Biden was less controversial but largely ineffective”. Conversely it implies “Trump tried to address some major issues, but failed to achieve unity.”
If one prompted ChatGPT re the premise that America’s successful future will rely on change in our economic structure, Trump policies upsetting institutions might be positioned as a necessary cost.
Conversely, if ChatGPT were prompted to focus on achieving more equal outcomes across different population minorities (including the white males), Biden woukd likely be viewed as more successful.
So IMO it’s the objective given in the AI prompt that matters the most. Which do Americans want more a) economic prosperity that lifts some more than others or b) equality of outcomes that tries to eliminate winners vs losers?
Yes, it's easy to bias a request as I did in my first ChatGPT query a couple weeks ago.
The punchline is now that you've published this, it is going to show up in other people's AI queries as an actual Wall Street Journal editorial
Patrick, you unfortunately may be right. AI generated news is quickly becoming a problem: https://www.library.pima.gov/blogs/post/a-particular-kind-of-disinformation-pink-slime-journalism/
It seems to have some moderate conservative bias and it completely missed the causes of inflation which was over spending on stimulus spending which was caused by Trump not Biden.
Basically December 2020 and January 2021 saw huge amounts of stimulus happen causing the situation where you have too much money chasing too few goods. No one saw the problems at the harbor which also caused normal shipping to slow to a crawl. It did not talk about the fact that an additional 400,000 Americans died because of Trump's pandemic response.
https://www.reuters.com/article/us-health-coronavirus-usa-economy/u-s-covid-response-could-have-avoided-hundreds-of-thousands-of-deaths-research-idUSKBN2BH1DK/?utm_source=pocket_saves . It also missed Trumps failure in World Trade.
https://www.forbes.com/sites/kenroberts/2021/01/14/trumps-trade-deficit-legacy-three-largest-in-history-in-four-years/
Finally the best and shortest synopsis of the difference between Trump and Biden can be summed up by the Economists endorsement of Kamala Harris on 10/31/24.
"While some newspapers refused to back a presidential candidate this year, today The Economist is endorsing Kamala Harris. Tens of millions of Americans will vote for Mr Trump next week. Some will be true believers. But many will take a calculated risk that in office his worst instincts would be constrained. We see that as recklessly complacent. By making Mr Trump leader of the free world, Americans would be gambling with the economy, the rule of law and international peace. Ms Harris’s shortcomings, by contrast, are ordinary. And none of them are disqualifying. If The Economist had a vote, we would cast it for her."
Moderate liberal bias
The poll does not appear to function.
This was my first poll, so I'm not sure why it's not working for you. I'm guessing only subscribers can take the poll--but not sure about this. I'll investigate.
What happens to the analysis if you add these last six months or so of Trump’s second term following all the recent decisions and outcomes
I am impressed by CHATGPT every day. Clearly the training data will influence the results, and not having access to that leaves one wondering. Thus it seems the answer can be predicted by the control of inputs as training data, so how does one make that unbiased and complete ? We all have friends that watch FOX all day, and we know how those conversations go.