r_hanz

This is exactly how I've achieved this on Windows in the past. However, since writing that solution, I have come to understand that utilizing AppDomains (for .Net Framework) and AssemblyLoadContexts (for .Net Core) to load/unload the binaries seems to be the intended workflow.

Very much this. One of the best accounts of this progression I’ve read is the “AI Superpowers..” book by Kai-Fu Lee. China held the “Knock-off/cheap copycat” moniker in the 90s, 00s, and early 1Xs, but the tides are shifting in terms of the quality of goods they produced. They’ve also taken a page from American business and begun outsourcing “cheap labor” to other third-world nations, in addition to playing loan-shark to the likes of Africa… You have to hand it to the way the Chinese have gained their footing in “capitalism” since the 1980s.

How much of that increase can also be attributed to distracted driving (ie driving under the influence of mobile phones)?

That show ages better every single day.

The title of this article made me think that paying down traditional tech debt due to bugs or whatever is straightforward. Software with tech debt and/or bugs that incorporates AI isn’t a straightforward rewrite, but takes ML skills to pay down.

Does iCloud mitigate this? It’s always confused me that iCloud is intended only as a data sync and not back-up… If one device goes down and the rest still work, could you still access data from the dead device?

Very nicely written article. Personally, I find RAG (and more abstractly, vector search) the only mildly interesting development in the latest LLM fad, and have always felt that LLMs sit way too far down the diminishing returns curve to be interesting. However, I can’t believe tokenization and embeddings in general, are not broadly considered the absolutely most paramount aspect of all deep learning. The latent space your model captures is the most important aspect of the whole pipeline, or else what is any deep learning model even doing?

I tend to agree, the flexibility that non-statically typed languages (i.e. Python) offer on smaller-scale projects (very) quickly diverges to chaos on larger-scale. With scale, rules and rigidity provide structure, without they provide verbosity and bureaucratic obstacles. Unfortunately “scale” is a gradient, not discrete, so there’s no “right answer” - hence the waste you experience. Ultimately, waste is in the eye of the beholder… “One person’s waste is another’s GDP.”

- The works of Richard Feynman: All are very nice reads. - The Idea Factory (Jon Gertner): The only book I’ve read more than once and gifted to several friends and acquaintances. - The Silmarillion: Incredible World-Building - The Master and His Emissary (Mcgilchrist) Dense, but rewarding. Considerably changed the way I think. - God’s Debris (Scott Adams, yes, THAT Scott Adams) Read it in undergrad and not sure if it was a JIT kind of thing, but it impacted me.

As a fan of Sinek’s messages - I would recommend his Ted/YouTube talks over the books.

Freemium version may get you left only… also pipe ads in. Sub gets you ad free (for now) and both buds!

If memory serves… the only difference is that one of the kernels being convolved is reversed for convolution.

Are you me?

Enjoyed the site personally when it was up, enjoyed this post even more. tyvm

I believe Nassim Taleb coined the term.

OP, my 2-cents having done this in a different industry within a company ~1/10th of yours: - as other comments mentioned, going slow and sticking to a solid vision of what “success” would look like in these efforts is important - most implementation challenges will be cultural, not technical - with the previous point, (and this may go against every other comment here) I have found it’s easier to teach someone software development basics than it is to teach the intricacies of your business/industry

with that said it has been my experience that a key first member in this is to find a passionate person within your org, whom you consider an expert or very knowledgeable in your business, and convince them to transition their full-time role to build these types of solutions. Find this person, convert them, and give them complete creative autonomy to pursue these efforts head-on. This person will implicitly possess an intrinsic understanding the problems with the highest value:effort ratio for the company, and when necessary, be able to fill the (many) necessary assumptions it takes to make functional apps that provide real value. You would be surprised at the simplicity and crudeness of what a smashingly successful app can have (and most importantly, what others will actually use) - so long as it addresses the right problem.

“Can its world model BE USED to produce…” “Can it BE USED to manipulate…”

FTFY

IMHO, the downfall is singularly: ad revenue. Monetizing your website/app/product indirectly in this way has been the sole source of many (all?) negative emergent features of modern software and products. i.e. a focus on attention in lieu of quality directly incentivizes utilizing things like dark patterns, addictive features, etc.

One of my favorite books of all time.. I have gifted around 5 copies:

https://www.amazon.com/Idea-Factory-Great-American-Innovatio...

Three books come to my mind on this topic: 1) Atomic Physics and Human Knowledge by Niels Bohr - and to think this was a series of speeches is pretty intense… still haven’t read it through to this day after ~7 years. 2) The Master and His Emissary by Iain Mcgilchrist- I’ve been pushing on and off since late 2019 and I’m ~80% through having understood ~75% and retained ~35% of what I read. It’s been a nice experience. 3) Infinite Jest by David Foster Wallace - got ~30% through and haven’t touched it in years.

+1 - went to pick up a copy, but they were all checked out.

To be fair, I don’t have any examples of “good practices” readily in-hand. However, I did try to address why I thought others were less impacted by this problem in the second half of my post.

Much the same way you can have all the best gear and still fail - Google’s primary strength seems to be the DeepMind group. I’m not affiliated with Google, but IMHO the reason they will slowly die is because their engineering culture has taken a backseat due to their broken hiring practices.

Bad hiring practices aren’t exclusive to them, but from all accounts it seems like their internal focus is on optimizing ad revenue over everything else. I could be wrong or misinformed, but it seems to me like they are playing the finite game in the AI space (DeepMind group aside) while FAIR are playing the infinite game.

*meanwhile MSFT are simply trying to buy their way to relevance (e.g. OpenAI investments, etc) and carve out future revenues (Recall) and Jobs-less Apple is building their trademark walled-garden (AppleIntelligence?). Although the use of unified memory in Apple silicon poses some interesting possibilities for enabling the use of sizable models on consumer hardware.

Overall it seems like “big-tech” is by-and-large uninspired and asleep at the wheel save specific teams like those led by Lecun, Hassabis, etc. not sure where that leaves OpenAI now that Karpathy is gone.