Occasional Thoughts on Climate Change

[TheClimateChanger]

16 minutes ago, TheClimateChanger said:

Sorry, I don't have time to write my own posts. And ChatGPT headlines get far more engagement than any I could create.

 

27 minutes ago, Cobalt said:

You are very committed to discussing all facets of climate change, and I admire your effort. However, would you possibly be up for moving away from AI usage in your posts? I can't help but see all the hallmarks of it throughout your Twitter page. A lot of it undermines your overall message, and I'm not talking from the energy consumption standpoint, but instead just the overall strength of the rhetoric you and consequently ChatGPT use. 

Don't get me wrong, though - I get where you are coming from. But I'm just doing this as a hobby and competing against a steady stream of dis- and misinformation. There are accounts that are actually PAID big bucks just to spread climate disinformation. For an unpaid hobbyist to compete against a career liars, AI is an absolute must.

[WolfStock1]

Wait - so we're basically having discussion with AI, with TCC as a proxy?

No thanks.

Can we perhaps start a separate "No AI" thread?

[Cobalt]

10 hours ago, TheClimateChanger said:

Sorry, I don't have time to write my own posts. And ChatGPT headlines get far more engagement than any I could create.

This is simply not true. ChatGPT is a far worse science communicator than many out there. Its writing habits are also quite predictable, to a point where a still-substantial portion of the general public can sense that the vibes are “off” and stop interfacing with the posts entirely.

 

My suggestion would be to narrow the scope of your content and try writing headlines more gripping and information-rich than what ChatGPT can do. It is very much possible, and the reward is that, with enough effort, your unique posts have potential to reach further than 4x the amount of derivative AI posts would be capable of. I guarantee it. 

[donsutherland1]

On 5/8/2026 at 5:22 PM, Cobalt said:

This is simply not true. ChatGPT is a far worse science communicator than many out there. Its writing habits are also quite predictable, to a point where a still-substantial portion of the general public can sense that the vibes are “off” and stop interfacing with the posts entirely.

 

My suggestion would be to narrow the scope of your content and try writing headlines more gripping and information-rich than what ChatGPT can do. It is very much possible, and the reward is that, with enough effort, your unique posts have potential to reach further than 4x the amount of derivative AI posts would be capable of. I guarantee it. 

There is actually one paper that suggests that AI-generated headlines may garner greater attention.

https://www.mdpi.com/2673-5172/5/4/110

This concerns me. My worry is that sometimes the actual nuance is lost in the AI-generated content. As a result, even as the headlines might be catchy, those headlines could be somewhat inconsistent with the content. That creates a credibility gap (which already occurs with "clickbait" headlines). I am not aware of research on that aspect of AI usage.

I remain an advocate of AI, but suggest that one should beware of its knowledge limits (pre-training) and working context. The former can lead to an inability to analyze novel situations, extreme outcomes, etc., e.g., it's no accident that physics-based models remain superior to AI when it comes to forecasting extreme events. The latter can impede the quality of the AI's output even as subsequent models have grown better at following instructions and understanding the user's goals/intent. 

Overall, AI can create stunning visualizations/infographics. I use it at work for just that purpose e.g., creating graphics for presentation slides. I still need to verify the output as every now and then, things need to be tweaked.  AI can also assess the strength and weakness of arguments and help identify "blind spots." I don't believe AI should be used to generate all arguments, much less as a substitute for human judgment. That's the case even as there are more advanced features that can also be applied, an increased capacity in AI coding (Claude Code, Codex, etc.) and sustained work that can be carried out through agentic AI.

Finally, these are my opinions. Others may well disagree with me.