mountainFrugal Journal

[mountainFrugal]

The order in which you've presented them isn't random but it's only important for people learning the thinking-mode fresh.

Yes. The circular way of working through it is a good way to put it. I view it as rapid zoom in and zoom out, but circular works as well. After reading through the Ozinga book on Boyd at Fort Dirtbag, his framework is just formalized way to do Bayesian updating. This is how I think about it as well (my training). Each node in the graph ("goal") has a probability of being there in the first place. Each edge in the graph is an intuitive probability. You could formalize this as a matrix of probabilities of the connections. That matrix shifts based on up to date information coming in from the environment, interactions with people, energy levels, caffeine levels, etc. As a daily example, many of the probabilities of both nodes and connections go to 0 when I fall asleep. These are never actually written out, but technically could be.

@bicycle7 let me know how it goes talking to your concrete thinking friend.

[mountainFrugal]

# WINS
Since I started working on bikes and bartending at the shop, I've earned $1,045.75, which averages out to about ~$25 per hour. I'm putting all the money I make into advanced bicycle repair classes. Initially, I was waitlisted for the United Bicycle Institute class in December, but I eventually got in! This has disrupted my plan to apply, table, or attend a zine fest, but it's actually the perfect time for me to use some of my accumulated vacation time. Depending on the type of repair, my hourly rate will increase significantly. I've worked out an arrangement with the shop owner to handle as many or as few of the more time-consuming repairs as my schedule allows, alongside my other work responsibilities. Interestingly, when I mentioned my desire to become the local suspension expert, two people offered to contribute to a GoFundMe for any classes I wanted to take. I won't be taking them up on this offer, but this community is amazing!

ERE Fest was fantastic. Echoing what others have already said, I felt completely relaxed and true to myself. There was no need for explanations or additional mask-wearing. One highlight was being bombarded by @mathiverse with questions about my background and whether I fit the ERE mold model they were developing. It may have been confirmed that I'm an oddball even among internet oddballs. . I believe I had a meaningful one-on-one or small group conversation with just about everyone there. DW also felt very welcomed and had many great conversations of her own.

# IMPROVEMENTS/INSPIRATIONS
Refer to the bike repair section above.

I read most of the Ozinga book on Boyd while military aircraft exercises were happening in the skies above. I'll pick this book up again next time I'm at the Ft. Dirtbag Library.

# NEW DIRECTIONS
After completing 1000 figures, I need to make them more expressive and less rigid. I'm experimenting with some cartoon animation techniques to achieve this. I don't want to get into animation, but they've mastered this.

DW and I had an art date night. We enjoyed some wine and made Linocuts. The next morning, I made another one, and we printed them on 100% cotton paper. We started selling them for $5-10 each (depending on the size) at the bike shop. This could be a good way to upcycle used paper by first using it for drawing practice, then shredding it, and finally making recycled paper prints. Feel free to PM me for pictures of our designs.

# DRAWING
Various statistics.

I've decided that my travel sketchbook will now double as a travel journal. I had been keeping them separate for some reason, but after taking sketchnotes for various ERE fest workshops in it, I realized that adding words enhances the observations.

I'm in the process of designing a logo for the bike shop. I've gone through about 40 ideas and shared them with the owners and patrons to narrow down the final direction. I'll complete this in October and plan to have stickers, hats, and t-shirts made. I'll receive a percentage of the sales. Essentially, the owner is supportive of me running a micro-business within his establishment (similar to the prints mentioned above). This also has the advantage of focusing my artistic efforts on one of my many interests: bikes!

I've been commissioned to create 3 t-shirt designs, each at $150. The basic ideas for each have been approved, but I'll make some refinements as initial ideas are usually not the best.

# OUTDOOR ADVENTURE
Various statistics.

I've gone on some group trail runs with @quadalupe, @mathiverse, @Jin+Guice, and DW. I love running with others. .

We're trying to organize a "mountain science" talk series at the bike shop. I'll be giving a data visualization talk alongside the local avalanche forecasters. I've written several blog posts on local weather and backcountry skiing in the last few years. The head forecaster appreciates my approach to visualizations (as seen on the blog), so we're collaborating to develop some tools for the forecasters and potentially some publicly accessible tools based on the visualizations I've already created using publicly available data.

To come full circle, I met the non-profit individual who commissioned the t-shirts at the drink and draw event I host. He introduced me to the avalanche forecasters after I discussed my previous blog posts with him. Always expanding that network. .

# WORK
I've written another newsletter. Mailchimp is somewhat bloated, and I hope it will eventually be entirely replaced by an AI version.

As part of a science journalism project, I'm profiling several scientists and their research. I plan to apply for a fellowship this month. I want to become even better at explaining complex research topics to a knowledgeable audience using words, illustrations, and data visualizations. My science writing mentor wants me to focus on scientific subjects that are far outside my current areas of expertise, which will keep my writing fresh since I'll be learning alongside the reader.

Next weekend, I'll be working on a full suspension rebuild, tuning, and mulleting (placing a smaller wheel in the back) to convert an all-mountain bike into a more aggressive downhill bike. The shop owner is supervising the project, knowing that I'm in learning mode.

Between my part-time work at the bike shop, art sales/commissions, and further cost-cutting through bulk cooking meals, I informed my current boss that I won't be renewing any contracts next year. While I was excited about leading projects/programs, they would all pay significantly less for much more work, additional travel, and more time on-site, among other factors. Academia is fundamentally broken. Goodbye, once again.

12-year-old me would just want to work at a bike shop and create comics. I'll wrap up my current projects and embark on the bike shop/art FIRE journey in July 2024. .*

# SOCIAL
On our way back from ERE-fest, DW stayed with her cousin in Bishop. It turns out that his best friend is a former MTB teammate from college. It's a small world! It was great catching up with all of them.

ERE-Fest - Yes, it fulfilled the social aspect as well. . The workshops were so enjoyable.

I helped my bike shop co-worker move a bunch of wood he had milled from the local mom-and-pop mill. In exchange, he gave me three top-quality live edge cedar slabs ranging from 8-10 feet. They are currently drying in the garage. I also cleaned up his off-cut cedar rounds while he had his trailer hitched, which amounted to approximately 3/4 of a cord. I split, moved, stacked, and tarped it in 3.5 hours.

* To clarify, I can do this right now and have been able to for more than a year. However, I enjoy working on the projects and with the scientists I'm currently collaborating with.

[mooretrees]

Yaaassss, another inspiring and exciting update. I’m consistently impressed with your ability to focus over time on your goals. It’s really awesome to have a window into your life.

Can’t wait to hear about your experience at the bike institute, it’s definitely something DH and I are thinking about for him in the near future. And if he does get a job at a bike shop, I’m hoping we can also do a similar combination of selling our coffee/candles like you are doing with your artwork.

[avalok]

Never ceases to inspire me how you are able to keep so many differing parts in your life in play. Every update is great to read through in this regard; thank you.

Migrating onto the bike repair shop next year is really exciting. Also think it is great that this is very much Freedom To thing from the projects you're currently working on.

[mountainFrugal]

Thanks @mooretrees and @avalok!

The bike shop has been really fun so far. I look forward to hearing about DH's bike shop journey. I have 3 additional data points that highly recommend the classes at UBI. I will let you know.

[mountainFrugal]

"When I recently hit thirty-five myself, I went through the unpleasant but irresistible exercise of summing up my career in physics. By this age, or another few years, the most creative achievements are finished and visible. You’ve either got the stuff and used it or you haven’t. In my own case, as with the majority of my colleagues, I concluded that my work was respectable but not brilliant. Very well. Unfortunately, I now have to decide what to do with the rest of my life." -Alan Lightman

Research career has peaked. Now what? DIY Tenure

[ertyu]

... the biggest issue for most athletes is you spend a third of your life not preparing for the next two-thirds. One day your entire way of life comes to an end. It’s a kind of death. You just have to go through it and figure it out.

Andre Agassi

[mountainFrugal]

Constant death and rebirth. I am all in. Moment to moment. Day to day. Year to year. Decade to decade. Lifetime to end*

Constantly molt the old you to develop into the present you to molt in the future again.

*insert spiritual paradigm here.

Add: This does not mean this is not terrifying in some ways.

[AxelHeyst]

Amor fati + infinite self-overcoming.

[mountainFrugal]

Amor fati + infinite self-overcoming.

Sprinkle in a heavy dose of Festina Lente and you have yourself a paired down mF philosophy. . Moment by moment, flow states, near future, zoomed out 3rd person perspective.

[mountainFrugal]

Happy Halloween! A short update. Trick-or-treating starts soon!
# WINS
A local artist and I are going to start a co-op style art studio/gallery business in the next town. We wrote a five page business plan to offset the cost of the rent of the entire space by subletting to other artists and hosting art events. There is a particular downtown property we are currently negotiating with the landlord on a commercial lease. Overall not very risky because we can easily cover the rent with savings alone and each have a dope studio space setup. More on this in the coming months.

We split a half pallet of 330 high density fire bricks with a friend. $0.89/brick. The bricks are 50-100 years old and are all stamped. Depending on the stamp, individual bricks go for 25-40 on Etsy. Lol. We plan to use them for a pizza oven though.

# IMPROVEMENTS - INSPIRATIONS

I am finding inspiration this month in the 100 Rabbits project (https://100r.co/site/home.html) posted a few times here recently and from the artist Geowulf.

# NEW DIRECTIONS
I found a shop in the city that sells zines on consignment. I am going to print out 5 copies of all the minizines (8 total) I have to sell down there for $5 each.

I am learning scribus www.scribus.net as an open-source alternative to Adobe In-design. I need to have more complicated print layouts for larger zines.

I have found some great online tutorials from concept artists talking about workflows that take source material, quick sketch them, and then combine in novel ways to build worlds. Cool stuff!

# DRAWING
I have adapted and implemented the concept artist's workflow towards my own zines and mini-comics leading to better imagination drawing through iteration.

# OUTDOOR ADVENTURE
Various stats. Lower than average running because travel to two meetings took up way too much time.

DW and I spent a lot of time on the MTB trails this month. She did multiple sessions on a wooden roller feature with a jump to a step down. After fully committing with speed, she nailed it!

I am doing deep dives into the physics of suspension systems for mountain bikes. However, much of the best source material is written in specialized motorcycle science/physics books that are not available at the library. I spent too much in this category. haha.
# WORK
I went to and presented at two conferences this month. One on AI applications and the other on general data science. The lines are blurring between these disciplines.

I did witness someone give a completely uninspired talk on generative AI made completely by... you guessed it... ChatGPT. The presenter did not even practice and tried to pass it off like this was a novel use of ChatGPT. If you have any creativity at all or have specific nuanced opinions/tastes/preferences about basically any topic then you should not worry about the bulk of AI use. It will fall into the same category as this corporate drone... uninspired. The problem is that they have never developed good taste by doing creative things so will not know when the AI is giving back uninspired BS. This is the same problem with using AI for writing code (at least currently). You have to know when it is giving back bullshit.

Fully moving towards bike/beer shop, art studio mountainHipster lyfe. Damn it is delicious.

Another commission in one of my styles of a friends pitbull mix. .

# SOCIAL
We are just finishing up #inktober over in the art MMG. We are still open for new folks. We do not talk about forum stuff at all because it is now a majority non-forum members.

DW and I did a trail work day, hosted pumpkin carving/chili last night, and are about to start sitting outside for the trick-or-treaters. This is the third year of DW's candy preference data viz project. Spoiler Reeces always wins. Pancreases always work overtime.

[shaz]

Is that a jellyfish attacking the hikers, or some sort of alien?

[mountainFrugal]

It is a jellyfish memory from my coastal trip in August. It is just overlapping memories in the drawings. Although the thought of giant floating in-the-air jellyfish is terrifying. Just enough of a nervous system to silently seek out hikers through a chemical signal (sweat? CO2?) to inject fatal stings. Be careful out there with the horses @shaz, forest jellyfishes abound.

[shaz]

I'm sure the horses are alert to the possibility of jellyfish! They are very good about noticing predators.

If the floating-in-the-air jellyfish seek out sweat or CO2 then runners have to be especially careful. Watch your back.

Did the ones you see actually have long tentacles like that? The ones I've seen were whitish translucent things with no tentacles. They were disappointing.

~Jellyfish heaven where jellyfish go to get away from Mormons and drunk mofos ...~

[daylen]

I did witness someone give a completely uninspired talk on generative AI made completely by... you guessed it... ChatGPT. The presenter did not even practice and tried to pass it off like this was a novel use of ChatGPT. If you have any creativity at all or have specific nuanced opinions/tastes/preferences about basically any topic then you should not worry about the bulk of AI use. It will fall into the same category as this corporate drone... uninspired. The problem is that they have never developed good taste by doing creative things so will not know when the AI is giving back uninspired BS. This is the same problem with using AI for writing code (at least currently). You have to know when it is giving back bullshit.

There is quite a bit of creativity involved in prompting the AI into giving different flavored answers. Seems like for at least the next few years a creative human + a multi-modal bot will be a highly effective pair in just about any environment that rewards productivity (e.g. any job). Co-pilots or specially designed interfaces for all kinds of different jobs are being developed as we speak.

I find the question of when a single AGI will outperform human+bot on an arbitrary task quite fascinating given that expert opinions are all over the place. Some think humans have a special sauce that is extremely difficult if not impossible to replicate (I use to think this somewhat more than I currently do). Many think no new tech is needed and scaling up existing solutions will be enough in the next couple of years.

I think some new techniques are needed but the foundations are there and perhaps by 2030 most jobs will be able to be automated though will not be because human communities choose to keep some humans in the loop.

[mountainFrugal]

@shaz there were ones with longer tentacles washed up on the beach. Most of the ones we saw while kayaking were moon jellies. I am always watching over my shoulder when running for fantastic creatures.

@daylen - I agree that there is currently a lot of creativity in prompt engineering and it will give creative people an edge. My example of the presentation by ChatGPT was interesting because the person presenting did not take into account audience (experts in the field), take time to practice (various parts of his delivery were bad), edit it at all (had 2 movie clips that were way too long, 90 seconds in one case), etc. He actually apologized while giving the talk that the video clip was too long. I am sure that eventually the AI will help a person like this give a better talk, but the fundamentals of a good talk are practice, transitions, stage presence, etc. These are all things that ChatGPT cannot give feedback on (at least for now). Humans have a special sauce currently because they have at least some awareness of the social/political context they are operating in locally or within organizations. AI currently does not have the nuance or an easy way to get that data. Maybe with further refined prompt engineering as inputs it will get better?

[daylen]

Yeah, to some extent gpt4 can be primed with contextual information to give more situational answers. You can ask it to give answers as if it were an entity with certain attributes (e.g. a particular human archetype, a dinosaur), or you can ask it to give answers for a particular audience with certain attributes (e.g. bioinformatics researchers, researchers with their families). Emotionally charging the prompt can also modulate the answers. You could say that this is for an audience sensitive to the negative effects of climate change for instance and it will give a different response. Multi-modal is extending this capacity by allowing some degree of triangulation between different semantic forms, allowing for more precise navigation of the cultural space.

[mountainFrugal]

Multi-modal is extending this capacity by allowing some degree of triangulation between different semantic forms, allowing for more precise navigation of the cultural space.

GPT4 has been fun to play around with. I have not done a lot of prompting for these types of things, but I am wonder how long before it takes into account these dimensions without having to ask for it? As with my example of the presentation, that person seemed unaware (culturally, of their audience, of themselves)...so how long before the AI can preempt a better answer to a person that is not aware enough to take all these additional factors into account?

[daylen]

The next generation of models will include more culturally targeted architectures that will likely give more context dependent outputs along with options to give up some privacy for more personalized responses. ChatGPT is trained on a very diverse data set and so simple prompts may by necessity give very general, average, or dull responses unless further specified. So either the human must be more specific or the architecture must be more targeted for the outputs to achieve more creative realism.

Multiple modes are gradually being integrated together and this has resulted in better spatial reasoning and so forth. Though, past a certain degree of integration you might run into more hallucinations or synthesesia where the modes cannot map neatly between each other resulting in non-sense (or perhaps creative genius depending on the edge case).

[jacob]

...so how long before the AI can preempt a better answer to a person that is not aware enough to take all these additional factors into account?

I don't think that would be superhard to establish a theoretical (Ti) model of people that more often than not is more aware [of how someone thinks] than the person is of themselves. (It's a low bar to meet.) It's possible to estimate temperament by "word cloud" and Kegan-depth by the complexity of the perspectives given thereby communicating in a way that resonates better than "random" or "average".

An extremely simple strategy would be to answer at the same readability level as the question. This only requires generating a bunch of different answers and selecting the closest match.

It would likely be even easier for an AI that remembers everything than a human who has to rely on some kind of Bayesian moving average/contra-indicators. My point is that it's possible to get pretty far without "getting to know someone personally" like the typical FeSi-model.

The limit might be the size or lack of sorting of the data set. There are already complaints that AIs talk like Silicon Valley types because those were the ones providing most of the training data.