gharman

Yes! Paraphrasing my childhood:

“Mom, I’m bored will you buy this a new computer game?” “Lol no! Go outside.” “Well crap. I guess I’ll have to write my own game.”

> My new suspicion is now that they didn’t got drastically smarter, but they got trained on the user input on the previous generations.

Yes - that’s one of the ways in which they got smarter. More data more better.

I presume also lessons learned re training procedures, algorithmic advances, but data is a primary lever.

Any (non?)digital media? I would be surprised not to see this exact dynamic on BJJ Fanatics. So easy to pick up instructionals with the constant stream of sales and discounts. Not so easy to spend the hours it takes to watch and digest one of those.

Commercial libraries = private library where you pay a membership fee? I’ve not really heard of that here outside of arguably university libraries and similar.

But SF public library has a couple dozen branches or more?

> (as an aside: bland food won't work well on someone who lost their sense of smell from long covid, and prison populations were heavily affected)

The consistent texture of the Nutraloaf may “help” here. My spouse lost her sense of smell (for other reasons than Covid) years ago and now relies heavily on textures for food enjoyment.

IDK when I hire, the first list (as might be directly relevant to the job) is checking a box, but the second list is what catches my attention.

Agreed there is a semantic argument here. I don’t think anyone is anthropomorphizing LLMs. I’ve had to adopt “objective” with one customer instead of “task” (my preferred default term) because their industry overloads “task” with something else. Now that said, as you point out, the real objective is next token prediction not “help the user with a truth bomb.” and it just happens to have emergent accidental usefulness (I think a natural bias from the training data).

They (chatgpt et al) do move closer to “clear and coherent” by adding extra layers such as a beam search on LLM outputs. Good to remember that ChatGPT et al are products, not bare metal LLMs.

My elderly father still uses this as his primary gateway to the internet and email. For all intents and purposes, this is the internet to him.

Looks like Scholar Turbo has GPT-4 and Chatpdf is still on the waitlist. But yeah, these chat-with-your-doc webapps are popping up out of the woodwork.

My long-horizon strategy has been the kinds of roles I take on. A product-focused role lets you go deep into a narrow problem area, whereas a customer-facing consulting/services role forces you to go wide, learning a new area every time. Over the years I’ve used time in the latter to select what to go deep into.

Indeed! I built a system just last year with - count em - three parsers to deal with PDF table extraction, including one built on TableTransformer. And then when GPT4 came out I just copy pasted a PDF into it as-is and darned if it didn’t do at least as good a job.

Now I can’t do this in earnest because of document privacy issues but I’ve diving down the rabbit hole of how small can we go and still get decent results. Spoiler: gpt2 is too small. :-)

> it’s 30+ minute charge time. As great as electric is, an extra hour in the car on a road-trip with 3 kids is a tough sell.

I dissent! Just did a 2-day (each way) road trip with young children and those charging breaks are pretty great and well-spaced for letting them run around and blow off steam. It’s not such a bad idea for the adults either…

IDK I’d say high school. I took a semester welding class at community college and we only spent 3 sessions on MIG (what this appears to be) before moving on to other processes. Most of the learning was around dealing with PPC, listening for the sizzle, identifying a good bead and common mistakes. This seems expensive for a one-class tool, especially since they still need the regular welders to complete the training.

You can choose to program functionally in Java or Python! I do this enough that seeing object oriented Python irritates me. I see this as more of a stylistic choice than forcing new languages.

Cost differential is an interesting thing to consider if this were done at scale. That single missile cost ~$500K (plus flight time for the F22 isn't cheap). How much would each balloon be, especially if some were decoys without any real payload?

That said, drones vs ships seems like a starker cost differential than this.

Funny thing is I’d bet the small amount trickling in to the long tail charities is more impactful to them than a larger amount to the big mega-charities.

I run one of those long tails and Amazon Smile was just right to cover running costs for hosting, bank fees, etc.

Make It Clear by Patrick Henry Winston has a permanent place on my bookshelf. He breaks communication down into a structured system and applies it to a variety of forms including (beyond standard prose documents) things like decks, pitches, and lectures.

I'll raise my hand as a data point. MIT, about 20 years out. It still seems to carry significant weight. I don't even work in the field I studied, so it's definitely not about any particular material I may have learned there. My conjecture is that association with the MIT brand implies that I must be a sharp cookie and someone worth paying attention to.

Not in the Kessel Sector.

That’s a feature! That 90yo probably has rather more worth sharing than the 18yo.

Like what? Serious question, not trying to be snarky. My understanding of these “nature vs nurture” discussions is that environment commonly refers to anything other than genetics - effects not “in-built” that could be caused by upbringing, society, nutrition, etc.

Yeah my first thought was “if only...”. Sadly, it’s still working just fine for me.

I did this but honestly in hindsight it didn’t tell me a lot. Two reasons come to mind:

1. Watching other peoples’ kids is a time-limited exercise. A major challenge with your own kids is that it’s constant and effectively without end.

2. There’s a big difference between kids and MY kids. I just have more intrinsic interest in my own. I do have more “tools” available to me to engage with others’ kids now (up to my own kids’ ages at least) but I care less. Not saying I dislike others’ kids, just that it’s different.

This reminds me of Snorkel (though unclear from the article if they’re using Snorkel’s trick of aggregating many weak heuristics). It can be made to work even in the real world. The rub is that coming up with these programmatic labelers is easier said than done especially for complex data.

It works well if a domain expert can say something without “cheating” and looking at the data like “put a box around round red objects because those are always apples”. But in practice people tend to cheat and look at the data first, and you end up with humans trying to emulate ML, poorly.

Who controls pictures of the grandkids, controls the eyeballs.

It’s percent of revenue over whatever period of time you care to measure it.

I founded, grew, and sold a boutique software services co. Here's the skinny:

* The core revenue model is charge customers a certain amount for doing work, and to have a lower internal cost. (duh)

* Typically you're looking for a 30% margin or thereabouts; after cost of sales, consultant bench time (yeah you gotta pay that even when folks are idle, if they're employees), etc. you have maybe a 10% profit margin when you're at decent utilization.

* Hourly and fixed bid are the primary structures - and I've seen plenty of both. This is simply a question of who will take on the risk of overrun. If it's the services shop (i.e. fixed bid), then they pay for that additional risk by padding the price above their best hourly estimate (that padding is what pays for the services if it does overrun, and is the reward for taking the risk if it finishes on time).

* You can lower hourly costs and increase predictability on your resources (an endearing term for employees/people) by hiring them full time, but you take on the risk of paying them for bench time (when you don't have a contract for them, they still get paid). You can avoid this risk for higher hourly rates by using subcontractors. A good practice is to keep a "bench" of folks in your primary areas of expertise and to supplement with subcontractors for speciality tasks.

* Ongoing maintenance revenue - maybe, maybe not. Wise customers will want to learn how to maintain their own software and include training as part of the contract - or even embed their resources in the team. This isn't quite like sales of a big enterprise software platform in that regard.

* A better long-term practice is to utilize the "foot in the door" to spread out and find other, new work within that customer's organization. Of course existing projects will continue as long as there's a need and the customer is happy, but I don't see a lot of "maintenance" contracts per se.