How ChatGPT and Google Gemini Can Kill Creativity

I’ve written enthusiastically about my use of Google Gemini as a kitchen assistant. And I’ve found a few other uses for it and ChatGPT in summarizing areas of research very quickly at high level. They miss a lot and are really bad at backing up their information with facts. But as a start, they’re useful.

It seems to me that the reason they’re useful in this way is that they are so median. That is to say mediocre. When you search recipes on the internet there’s huge redundancy with bad ideas sprinkled in. The LLMs ignore the outliers and give you a nice middle of the road consensus. Which is my personal method- I look at a bunch of recipes, averaging mentally over the proportions and list of ingredients until I have my take for the moment. It’s easy to do this with an LLM, asking it to take out the olives or asking whether lemon juice would be a good substitute for the vinegar. And these models can be quite opinionated if their training set is opinionated. Of course some of those opinions are wrong (don’t salt beans during cooking), but useful.

But I made the mistake a week or so ago of asking them to help write a Substack post. I had a page of text notes and an outline. So basically all the ideas I needed to start the through composition of a first draft. So I thought, why not give my notes to Google Gemini and ChatGPT and skip that first draft?

So what I got was totally, as the kids would say, “mid”. I mean it was what a mediocre thinker would do with my notes. It put in all kinds of caveats and “as if” statements to route around my unique take on the relationship between brain and intent.

Not only did it water down the ideas to non-existence, if I tried to edit both or their essays back to my liking, it was like finding I had a set of false beliefs, as if an alternate universe version of me had written something I disagreed with.

I had to erase their efforts, take a walk, and come back to my notes and do that first draft. I’m not sure the product was the same as if I had never let those things near my work. So not only does the LLM flood threaten to dilute the content of the web, it may well threaten our ability to hold opinions far from the median.

In finishing up my manuscript and starting these Substack essays, I’ve realized that my way of looking at being human is now pretty far from that median. I’m in the midst of reading Anil Seths Being You and from the first page, I find the approach to be unhelpful. This idea these academics who study consciousness are stuck in a false dualist “mentalism” is becoming more clear to me and will probably be my next series of essays over on Substack once I get through the current set of ideas on Self and the Power of Pretending.

Week 3 Substack

I’m on a three week long streak posting an essay on Substack. This week, I’ve extended the conversation about self and values to where values actually come from.

These posts are experimental, but at least being read on the site at a pretty steady rate. As I suspected, like other social sites the algorithm is putting my writing directly in front of some readers, something that is hard to do here on the blog other than through search engagement. But I have no plans to give up my casual journal efforts here. I added a Substack subscribe form to the site here to push my readers here to sign up for the posts there. I’ll probably archive them here at some point, but for now the idea is to see whether there’s really interest in reading my contemplations on brain and mind.

The idea on Substack is to keep to a publishing schedule and put out something of a complete thought in each. As I’ve written here, it’s a drawback of the blog format that it promotes this kind of running narrative and makes it difficult to communicate more complex, bigger ideas. The truth is that really only book format can be structured and long enough to really allow for exposition and elaboration of ideas. But I think that I can at least get better at explaining and chunking down ideas by regularly writing these explainers for the general audience on Substack.

Plus, it’s the start of the next intellectual journey. Or at least documenting where I’m going since slogging through turning “Deciding Better” into a book now that I’ve been at it close to 25 years. I’ve ended up in a place that I now realize is both in the mainstream of thought among neuroscientists, but pretty shocking to anyone outside of the neurosciences. It started when I began to realize that when we feel like we’re deciding, we really are more realistically being aware of what the brain is doing and if we wish, probing and rationalizing the choices we’ve made. The book is all about how to improve the process, but it can’t restore the illusion that there’s an “I” in charge of the brain. It’s quite the other way around.

So no wonder that in these first essays on Substack, I’m examining things from the other side: What is the “I” that I think I am. How is it constructed and to what extent can I control or change it. Plenty more to come.

In Suspense

I appreciate the comment on my post election thoughts from Dave at Nice Marmot, saying:

What James suggests smacks of “both sides” equivalency.

This is something I’ll freely admit to, but only as so far as trying to meet people where they are. My life is very oddly divided between members of the two sides. In fact, today, among the Trump supporters, the vibe was very much, “We won the Super Bowl. Season’s over and we’re ready for the next fight”. Which just reinforces my hypothesis that so much of this division is driven by group identification and cheering for one side over the other.

I know one thing from personal experience. It’s fun to deflate false views by logic and fact, but it doesn’t make you any friends. In fact, I generally see it harden positions as it ends up personal attack. I still remember getting upset at someone during the COVID pandemic who was in a group complaining about the truth about Ivermectin was being suppressed by the authorities. I guess I took it personally as a physician, but to this day, my interactions with that person are not the same, even though he was only repeating what he heard in his environment.

I had suspected that the outcome we got was more likely than the polls were showing and defended Nate Silver’s complaint about pollsters “herding” results to suppress outliers, which should be more common than reported just by chance. Mostly, it was a combination of the reaction to social change I kept hearing and the economy of the Biden presidency with its inflation. I don’t think that the achievements were really ever communicated. Per this nice summary: Bidenomics Is Starting to Transform America. Why Has No One Noticed? – The New Yorker. In fact:

“By a steep margin, Americans did not approve of Biden’s presidency. By an even steeper margin they thought the country was heading in the wrong direction. They were demanding a new direction that Democrats never figured out how to offer.” Washington Post

So now we move from uncertainty to suspense. I’m following Dave’s planned move up north and wonder whether he may be a growing trend of regional migration where we cluster by economics, education and social attitudes. While I had expected demographic shifts to swamp regional divides, it seems like our system is also promoting localization among the group because of our non-proportional systems of governance.

Ecological Post-Election Thoughts

From an ecological or systems level point of view, one can look a bit beyond the politics, human rights and economic effects of this election to understand how these things happen. I will say that it troubles me that we have half the country looking at the other half of the country saying, “I can’t understand how you could have voted for him/her”. And really meaning it. That they cannot summon an empathy, a theory of mind that lets them imagine being someone on the other side.

So first, as a human behavior, this is quite typical. To trivialize it in the extreme, the Yankee fan has no understanding of how anyone could be a Mets fan and that Mets fan (my family’s group identification) could not understand how anyone could cheer for those Yankees. In fact, some anthropologists have suggested that the more trivial or untrue the group beliefs are, the stronger the group affinity since facts can’t possibly sway a group member into defecting.

So we choose (or are born into) our group and that identification leads to our rationalization of our loyalty. And as long as beliefs aren’t upset by what I think of as “brute facts” that make belief untenable in the world, we hold those beliefs tighter and tighter.

So let’s just say that voting is a simple application of group identification where we try to figure out which party is aligned with our group identity. Some identify by party, some by creed, some by social standing, etc. Without going into detail, I believe it’s clear which party captured a bit more group identification this time around.

The second factor is our regional single representative by majority system of government. Which is opposed to a proportional system where groups are allowed seats in government based on the percentage of their support. In political science, “Duverger’s law” says that this type of single member district election will lead to a two party system. As I envision it, basically, every group is trying to be on the majority side because only the winners get to govern. So it tends to lead to a 50-50 split where anyone not feeling represented by the majority will defect to the other side to gain influence there and build a majority. I believe this year we had a lot of defections by those who felt they were not being heard by the side in power and put the other side into the majority.

It’s important to realize that this group identification and picking sides in an election is by feel. One side feels more right than the other. That’s why all of the arguments and fact checking don’t lead to a rational decision based on Multiple Attribute Utility Theory (MAUT). And even if you ask how the decision was arrived at you’ll get a rationalization or an answer representative of that group fit with one side over the other.

It’s a process of alignment and realignment. And a losing party can’t steer to the correct side by analyzing previous outcomes because the many unspoken factors that provided that comfort with the other side can’t be fully surfaced because they vary for different individuals with their multiple group identifications. Making it even more difficult some of the factors either aren’t known to the individuals themselves or can be recognized by them but culturally must be left unspoken.

I’m no political consultant, but I’d really like to hear more empathy and understanding. To my way of thinking, honest empathy is the only way to improve our ability to engage others productively rather than simple cheering for our side and insults for the other side.

Reading List Updated

It’s getting towards the end of the year and I noticed I’ve neglected my Reading List page for this year. So I spent some time catching up.

It’s interesting to look back and see how I’ve divided my time between genre fiction, literature and non-fiction over the year. Lots of audiobooks this year, slowing down a bit in the second half once I’d run through Stephen King’s, Blake Crouch’s, and Hugh Howey’s big works.

I’ve long called this website by personal journal and the intellectual journey is well documented by the reading list. I’ve become interested in how we come to construct our beliefs and our perception of self by reading books in both Neuroscience and what is basically the anthropology of religious belief. It may be that as my intellectual memoir about how I came to understand how people make decisions that my next will be about coming to an understanding of belief and construction of self. It’s an idea that was central to Bateson’s understanding in The Ecology of Mind (see here and here).

One of the reasons why I caught up with the Reading page is what I think will be a need for an annotated bibliography for the manuscript in progress. Plus, as I map out some Substack topics, I realize that, although I see that I’ve mentioned these ideas here over the years, I’ve never provided simple exposition. This is something I hope to remedy over time.

Substack- So Far So Good

A week ago I posted for the first time on Substack and result has been quite nice. More reads of a post coming out of nowhere than I get here on my 25 year old blog based on analytics, but of course RSS reads these days are silent, so total audience size of a blog is an unknown. The community at Substack is really nice, reminiscent of the early days of blogging but enriched with many serious creators. It seems like the audience is looking for engaging content.

Writing on a schedule is a challenge that I’ll have to work on, probably by prepping more than a week in advance. At this point my goal is to see if I can get better at explaining my take on the topics where neuroscience and life meet. So I’m trying to take a single one of the concepts each week and try to make it understandable. So last week it was the illusion of self. This week it I’ve written about externalization of mental events as feedback.

In the meantime, I’m talking to some publishing industry pros about what to do with the manuscript. At this point, all options are still open, but I think a plan is taking shape. This writing, like my photography is my art. It’s for me and it’s my pleasure to share it.

My New Substack

When I saw that Venkatesh Rao, was “retiring” his blog of 17 years, Ribbonfarm and moving to Substack, it solidified my plan to try building audience there as well.

Here’s my first post there: Forget Cognitive Dissonance.

I’ve been writing in this space, hosted one way or another for just about 25 years now. The early days were exciting and we built up each other in a blogging community for a few years. Then with the rise of social media and commercialization of websites with ad driven clicks, blogging died a quiet death for all but a few high profile public intellectuals. I get traffic here from what I think is a small group of regulars plus some pages that Google serves up in the first page or two on note taking tools and camera equipment.

At least for right now, Substack feels a lot like a 2024 version of that early blogging environment with easy interaction with writers and low barrier to entry. Of course the monetization is a big problem with the balance between free and paid content varying wildly between creators.

For now, I wouldn’t think of charging for content as I’m after some more engagement than one gets at a site like this. I’m putting myself on a once a week schedule there with. more frequent notes here.

My other motivation, besides engagement, is to work on writing for a general audience to develop the ideas I’ve put into my manuscript. That’s put aside for the moment, but I know that the next edit should be an effort to be clear and systematic in exposition of ideas. I’m also curious about what subjects a general readership is interested in besides tools for thinking so that I can focus my next big writing project on something that might be more broadly read.

Blog Engine Running: Tools for Deciding

Dave Rodgers always notices when I start feeling the need to write in this space and I thank him for that. I’m glad to know that there are a few consistent readers.

A little while ago, I wrote about my review of what actually gets read here. It’s not the book reviews or philosophy, it’s the discussion of tools. And that makes sense, because it’s by far the most useful information for the largest audience.

And the last section of my manuscript is about positive steps and techniques for making better decisions. So it makes sense to focus there while trying a bit harder to lay out the unique perspective I’ve gained by looking into the brain mechanisms behind our choices. Because it turns out that it’s not so easy to make better decisions when almost all of them take place entirely outside of awareness by processes we have no access to or influence over.

Second Draft of ODB Manuscript Done

After many false starts over the years, I started from scratch and finished a first draft detailing what I’ve learned over the years almost 3 years ago. I let it sit for a long time, but after picking it up again about 18 months ago, I’ve made it through a full reorganization and revision and have something that more resembles a book.

Along the way there were a few long detours into researching probability (focusing on Jaynes’ big book and) Polya’s two part treatise on plausible reasoning) and the neuroscience of perception and valence. So it’s about 75,000 words. A sizable book, but one that’s not really commercially viable in today’s competitive publishing world.

Why would it be tough to get published? Two reasons really. One is that I’m not known in academia, philosophy or as a public intellectual. This little corner of the web is visited occasionally and I’ve never built an audience eager to spend a ton of money on a book. Given the investment a publisher makes in preparing for print and distribution, it’s not a reasonable business proposition. Second, the subject of decision making is a very crowded area in the publishing space. For the most part, Economists have flooded the zone with their discussions of “irrationality” and “bias”. So books on choice and rationality abound, even if they don’t really help us much since most of our decisions are made outside of awareness and can’t be adjusted to avoid our built in biases.

It’s good to have this big project behind me and begin to think about next steps. As I’ve written recently, I’m look at options and but I think there’ll be more effort here as part of the path forward.

What Works On This Blog

L1001460_24-09-26_LEICA M11.

Thinking about how to best use this blog.

When you look at my most read posts here you’ll find it’s all tool posts. The Hobonichi, the Plotter, Cameras, etc. It’s true of my blog reading habits as well. Reviews of software, stationary, cameras, etc. I have to presume this is universal since the tools cut across all of the other areas of interest which are much smaller- topics like decision theory, systems and neuroscience. Yet we know that these small niches are where dedicated audiences are built through consistent posting. I read blogs on AI safety, quantum mechanics and statistics. All of interest, but nothing directly useful to me.