logiduck

How do you "translate" music into vibrations

Is there anything to translate? Music is vibrations to begin with.

I have had exposure to ABA but mostly through the lens of the parents with children using ABA and it was mostly positive. Being able to take their kids out and put them in the car for the first time without a breakdown after starting therapy was a major achievement in their eyes.

I am just curious though about what you limits are? Isn't everyone being put into a box? Isn't that just being part of society? I don't know what your experiences are, but probably isn't there a spectrum of ABA from good to bad just like there is a spectrum of all types of interventions and parenting from being overbearing to too lenient?

Just genuinely interested in that it seems like being part of society unless you are part of the 1% has a big aspect of conformity and "fitting in" to society even if that isn't what you want to do. That historically to me has just been known as growing up.

There is a lot of money from Silicon Valley rushing into ABA treatment and having seen the results I can't help but think we do better.

Current systems of ABA rely on billing the therapy to insurance and thus most of these ABA "mills" just throw inexperienced college kids who haven't graduated yet at the problem on a session by session basis paid at minimum wage.

These college-aged therapists have no interaction with each other since they go directly to the session at the client's house, leave and go home to take notes. Sometimes they switch patients quite regularly so there is never a chance for a comprehensive plan to emerge.

The insurance only pays for treating the patient and thus no space is available for training the parents. If we want to do better, we have to start with the parents and develop therapies that last the whole week not just 2 hours a week.

Theres a lot of potential for improving the system, but the current ABA system that SV is taking over focuses on extremely short sighted "box checking" and does nothing to develop a long term approach for the family.

Yes, that is exactly my point. You will have to rely on AI to translate it back out, but that translation is built on probabilities not machine rule-based translation. So you can ask and have the AI explain everything to you, but you are still trusting the "black box" to tell you what is happening. Very different from today.

Also, you can "get good" at reading assembly, but that doesn't matter if the AI can output a custom OS from scratch and a custom VM to execute the program it wrote to solve your use case. It will be so impossibly complex that it would be the equivalent studying protein folding.

Instead people will just trust the AI.

It also won't help you if the code base the AI produces for a SaaS app is a million lines of assembly.

Instead of having different layers of OS, compiler, high level language, an AI will just be able to produce one layer. because after decades of trusting the AI to write our code, why wouldn't it?

The current gen of AI outputting code that in human-centric programming languages will be a blip in the history of AI. As it advances, it can just skip that step.

Its will be orders of magnitude more complex and opaque than anything we have today.

Not really at all.

What you are describing is an traceable transformation of code in which there are several intermediate layers that people can inspect and understand. They can inspect the exact rules for that transformation. The process is repeatable and verifiable.

What I am describing is a black-box stochastic generation of low level code in which there is no higher level representation anymore. AI generating Assembly not by a set of rules, but using statistics. There will be no individual layers to unwind or inspect, because for AI it doesn't need them. Our separation of concerns was built for our human brains and limiting complexity of projects to our understanding.

We may have the tools to manage it, but we are losing the ability to understand it.

AI writing software will be a exponential explosion in software complexity.

AI would very well create its own programming language to be more efficient for the AI to code in that we have no hopes of understanding. Imagine that AI started to output a large SaaS app written today in Python in Assembly because for the AI the extra cognitive overhead of using Assembly doesn't exist. At first we might resist and tell it to use a language we understand, but then as time goes on we grow more comfortable that it does the "right" thing and down the road people are just generated raw Assembly using AI without really understanding what the code is doing, only looking to see if the code behaves the way they expect.

Imagine entire codebases spun up in seconds with so many lines of code, a single person would never have hopes of understanding everything, needing to rely on AI to summarize and explain the code for them. Now imagine that massive code base being iterated and worked on for a decade over the life of a company.

AI could bring Terabyte sized code bases in a decade or so.

Copyright laws do not support your argument.

There have been many cases in music where the offending song was forced to pay because it was "close enough" to the curve but not touching it.

Of course there is fair use, etc etc. But modern copyright was a reaction to the printing press so that people were motivated to still create content in the face of new technology.

If current copyright laws do not protect people from creating something new because of fear that they will be ripped off, then new copyright laws need to come.

One thing I wondered about how arguments from one side of the issue say that AI copying and extracting information for free isn't stealing, but what if you use their argument against things that aren't copyright like secrets, military, corporate, trade secrets.

Like can if an LLM saw the coca-cola forumla and the weights are released, what are the consequences? If it ingested top secret confidential information and released the weights, I assume that counts as stealing something and distributing it.

Yes, this makes it very difficult to apply ML and RL in non-simulated scenarios.

With simulated scenarios you can just replay and "sweep" across hyperparameters to find the best one.

In a realworld scenario with limited information, fine tuning hyperparameters is much harder as you quickly find yourself in local maxima.

Seems like the Moat they might have is licensing is if they start getting celebrities or studios to license their tech. Very big market for animation studios to have this tech.

If they were the shop where you go to get a famous person's or character's voice I could see it reaching this valuation.

But as a pure tech play, seems likely that in 2-3 years the open source models will be good enough to not need an established player.

If you go on Fiverr and ask for a text narration, most of the people on there just use elevenlabs or a similar service and aren't actually doing the narration themselves but essentially just licensing their voice.

I just saw an TurboTax Ad where a guy was like "I Like free stuff" and then it said he was "happy to read the disclaimer" on TurboTax and see that "Roughly 37% of taxpayers qualify" which he looks thoughtfully in the distance and says "Thats me!"

I thought it was a funny commercial because 37% doesn't seem like a lot and Turbotax is portraying it as the average person will identify themselves as part of that 37% even though that is not too far off form just 1/3 people so a minority of people.

It was one of the few times I saw a company blatantly lean into the negatives in their fine print and just outright tell you its good.

https://www.youtube.com/watch?v=iijnr4UR4QE

Ever since I was a little kid I was always hearing those inspirational stories about how someone would get into an accident and then "The doctor told me I would never walk again" or "I wouldn't make it to my next birthday" and then the person made a miraculous recovery and "proved them wrong" or something along those lines.

It always seemed like an attention bias to me. Because for sure there are many more stories about the doctor being right and the person did never actually walk again. And we don't really pay attention to the stories about the doctor saying they would walk again. Those aren't interesting.

It can also be a bias with the triumphant person overestimating the negativity of the past. Like did the doctor really say you will never walk again, or that it will be tremendously difficult and unlikely?

People like an underdog story and will hold it with more attention than other stories in which the expectations met reality.

The reason it exists is because of the products they haven't released yet. Spatial audio is part of the AR experience they want to release in the future, and it just so happens they can release it before that is ubiquitous so now they have a testbed for it with existing devices.

Its one of those things that you might not go out of your way to get, but once you have it, its pretty nice. Not game changing by any means, but it adds a bit of immersion.

Also, think of it as not an intentionally developed feature for the current gen of devices, but rather the "next gen" of AR glasses and such. I don't think apple went out of their way to build it as a selling feature today, but as a consequence of building it for their AR vision of the future, they can release and test it today.

It seems like personal tutors will actually be in higher demand if AI learning takes over. I mean kids have had an ever increasing array of ways to learn and supplement materials over the past 20-30 years and if anything education has been getting worse in recent years with many benchmarks for education reaching decades low levels.

Also, just a random thing that I thought of playing around with it is a few days ago a guy posted about an AI quiz generator for education.

If you ever need to pivot, it seems like this would do reasonably well in the education space also.

Cool.

Just as some feedback I did the demo with the "VW Beetle" topic and one of the test cases was:

Question: How did the introduction of the Volkswagen Golf impact the production and sales of the Beetle?

Expected: The introduction of the Volkswagen Golf, a front-wheel drive hatchback, marked a shift in consumer preference towards more modern car designs. The Golf eventually became Volkswagen's most successful model since the Beetle, leading to a decline in Beetle production and sales. Beetle production continued in smaller numbers at other German factories until it shifted to Brazil and Mexico, where low operating costs were more important.

GPT Response: The introduction of the Volkswagen Golf impacted the production and sales of the Beetle by gradually decreasing demand for the Beetle and shifting focus towards the Golf.

It seems that the GPT responses matches the expected but it was graded as incorrect. But it seems to me the GPT answer is correct.

In fact a couple of the other answers are marked incorrectly:

Question: What was the Volkswagen Beetle's engine layout? Expected Answer: Rear-engine, rear-wheel-drive layout GPT Response: The Volkswagen Beetle had a rear-engine layout.

was marked as incorrect.

For the chevy tahoe example, you are referencing the dealership, but in that case it wasn't a case of the implementation failing to do a positive test for fact extraction, but to test the guardrails.

Aren't the guardrail tests much harder since they are open-ended and have to guard against unknown prompt injections and the test of facts much simpler?

I think a test suite that guards against the infinite surface area is more valuable then testing if a question matches a reference answer.

Interested to how you view testing against giving a wrong answer outside of the predefined scope as opposed to testing that all the test questions match a reference.

Yeah, thats been my experience with it also. It works great for small things and getting going, but eventually the project starts to grow and you want to embed something that would be straightforward if you used a JS framework and then the spaghetti code starts as you bolt on a JS component inside of HTMX. After the 10th time you do this the project gets pretty messy and you want to just start from scratch with a conventional framework.

In addition the policies while they may not have matched what was necessary where in and of themselves a form of communication to get people to take it seriously.

There might not have been a reason to make everyone stay home, but if the only way you can get people to take a quickly evolving situation seriously is to over correct, then that is the tool you have to use.

By making the same rules apply to everyone and having such extreme responses it reduced the ability for the majority to say "yeah, its bad but it doesn't apply to me". By making everyone wear masks, it was the only way to get the people who really needed to wear the masks to do so.

The policies were extreme because the situation was extreme. People now have the hindsight, but I have lost at least 3 family members and 5 neighbors due to covid and the secondary effects. 4 of those died of ordinarily treatable causes (heat attack, pneumonia, etc), but the health system wasn't able to provide them the treatment needed because of Covid was overwhelming them.

I mean this coming from the best way possible with no ill intentions, but you really seem to be missing the mark by releasing a product with very thin utility and "writing" blogs that are surface level GPT fluff.

I wish you all the best on your journey, but it really doesn't seem like you are steering the ship at all. I understand you have the best intentions but the work that you have produced seems to be the very thing you are trying to avoid.

For example,

The AI quiz generator is designed to supplement educators' existing strategies, offering a time-efficient way to create diverse and challenging material.

An AI quiz generator is very basic rudimentary thing. It is much harder to actually design and implement an educator strategy to efficaciously use AI. You seem to be outsourcing everything to ChatGPT while making no efforts at solving the real problems; the very thing you complain about.

I hope you take this with my sincerest feedback, but you need to do the hard work. Not just generate vapid article with chatGPT and say you are doing research. If you are doing the research, your work that you are presenting doesn't reflect much effort.

It's really bad at the university level here at the moment, feels like we have a loop of ChatGPT producing learning material, students submitting work written by ChatGPT, and then lecturers marking it with ChatGPT.

Currently, I have a quiz generator hosted on [redacted], it uses OpenAI API to create a quiz on an input topic or based on uploaded PDFs

I guess I'm confused on your intent here? You are complaining that things are really bad with everyone using AI and your response is to build and release another AI tool?

everyone seems obsessed with the short-term quick wins, and being fresh out of the UK education system

Seems like an AI quiz generating is a short-term quick win.

I was hoping when i first started reading your post that it would be pedagogy based around teaching to learn without AI and how to use AI in the best ways. Maybe I am misunderstanding the project, but seems like another cheap short-sighted attempt and getting kids to regurgitate content in the least meaningful way possible.

Edit: Reading your site, edu-ai.co.uk not sure if this is a troll post or not, but looks like every post you have there is generated by ChatGPT.

yeah, in a major metro area, I get most packages next or same day for free pretty consistently.

Outside of the US, (Latin america, europe, asia) it is more common to have an all in one usually in the kitchen. A lot of the foreign workers who stayed in corporate housing we had didn't even know what a dryer was or how to use it.

doubt that headphone can make your body feel the vibrations also. Listening to music is a full body experience not just through the ears.

Why is it sad to pursue something that will never reach perfection? Why is it sad to go on a path literally thousands of people have gone on and have done better than you? Release yourself from these things and you will be much happier.

Music taught me you don't need to be perfect or the best. Music is about connecting with other people. Its much better than computers for that. Its a very "meatspace" realm and while the virtual space has made music more accessible, it will always be firmly fully experienced in the in-person physical world that requires another human being.

What else are you gonna do with your life, this is it

Well i mean there are tons of things out in life that you can do besides sit in front of the computer producing code that very few people will see.

I'm not sure how this will resonate here, but this past year my interests in side projects sharply declined. For reference I am early 30s, married no kids. Up until now I used to work on 2-3 large side projects a year, diving deep into them obsessed with sleepless nights working.

Now with LLMs it just feels kind of like whats the point? Either my work will be consumed by an LLM to train and get tossed aside or i should just wait 5 years and whatever i will have worked on will probably just be a prompt away. Even if that doesn't become true, nobody really cares about traditional software building anyways. Everything is just LLMs. All interesting things about AI that I really liked building with ML or RL now just seem completely obsolete. Any mention of AI is just overwhelmed by "oh so your just connecting to ChatGPT and boom no problem, right?" with the majority of the population completely blind to the fact that there are many types of ML and other things besides an LLM.

Its just hard to get motivated by anything. Unless you are working on a LLM right now nobody really cares what you are doing in software. Just seems futile.

I think also my own personal expectations get in the way of doing side projects. Its hard to work hundreds of hours on something knowing the end monetary value is going to be $0. So I end up in this state of wanting to work on something but then getting cynical, realizing there's little to no money to be made and just demotivate myself.

Anyways that is a long rant and this past year instead of building software i started to learn an instrument. it has been great. with software there is this notion of something being "optimal" but with music there is no optimal. Even the highest level artists are not satisfied with their work. The music is always above you.

The core issue of the secondaries market is that companies have a conflict of interest not to participate and the control that they want in order to participate essentially violates securities law. Founders want control over the price, who buys the shares, who can sell, etc. Carta and other companies so far haven't been able to solve this problem on mass scale of motivating founders to allow all shareholders to sell on an open market. There was weak demand from the company side so it seems like Carta tried to force it by making the market first and then asking for forgiveness because asking for permission wasn't working.

As a founder it is easy to see how the actions of Carta seemed like a breach of Trust. but the lack of a major player in the secondary market is a problem for past employees of startups who can't sell.

Putting aside using data they shouldn't have used, this episode is revealing a major problem in the startup space.

Unfortunately there is conflicting interests when it comes to secondary liquidity in a company. The founders/CEO wants control of the cap table, who the investors are, and the share price. They also want any interests in purchasing shares to go towards issuance of new stock in a round rather than cashing out past investors.

But for employees and other investors, secondary markets are a positive thing. If you are an employee sitting on millions or hundreds of thousands of dollars in shares, it can be frustrating that the company locks up and controls all liquidity even though you might have paid considerable exercise costs for those shares. As an past or current employee it should be a positive thing to have inbound demand for your shares, but this goes against the interest of the company.

The reason Carta's secondaries business failed is because founders are not motivated to provide liquidity to employees or investors. Founders and the C-suite want control and typically can self-approve transactions for their own liquidity. This conflict of interest needs to be solved by someone.

Hopefully someone can solve this problem in the future.

Seems like a stretch of the imagination. VC funding is already bone-dry so there are about 1000 different reasons not to get funding as it is. VCs are only investing in strong conviction plays. I'm skeptical that an investor would have strong enough conviction to want to fund you and then back out if you use Carta and put that in writing as the sole reason for not investing.

Even then the onus is pretty big for demonstrating damages for lost opportunity rather than direct financial losses. For example if you called your broker to put in a trade and they missed it and the stock 10x the next day, your SOL even though there is a theoretical missed financial windfall. its a different story if you asked them to close a position, they didn't and then you lost money since the latter actually has tangible losses as opposed to missed theoretical gains.