Letting your JS/TS compiler convert it into supported form. Not really a polyfill, but it allows to use new features in the source and still support older targets. This was done a lot when ES6 was new, I remember.
There seems to be no stable consensus on which LLM one should have used, to get good results. Which is somewhat natural, things are moving quickly - and evaluation methods are immature (and the little we have, actively gamed).
But a lot of the arguments seem on the surface to be of "No true Scotchman AI" form. Or "you are just holding it wrong" (ref Apple).
Sweden as an example is 10 million people. The world is 8000 million, 800x larger. The affluent parts of the world are somewhere between 100-1000 million, so the market potential is 10-100x larger. If one is able to address the global market, the economy of scale around manufacturing will be heavily in favor of that. This is particularly true for hobbyist stuff, because it tends to be both (relatively) low volume and low margin. And also price sensitive, in that people might just adjust their hobbies based on what is affordable/not.
This is just demand side. Of course producing in Sweden will be more expensive than in China - for electromechanical things, Shenzhen is likely better on every single metric...
RF anechoic chambers are, as a side effect of their construction, pretty low echo also when it comes to audible frequencies.
I have spent a bit of time in one (EMC testing a product), and it was the quietest room I have been in by far.
If investors think that the acquisition will increase the valuation of the company by 3B (or more), and the acquisition is paid fully or mostly in stock - then one can it basically for free.
At least having good integrated support in the package manager for an allow-list of packages would be good. Then one could maintain such lists in a company or project. And we could have community efforts to develop shared curated lists that could be starting points.
If that really catches on, one could consider designating one of them as a default.
Might also want to support multiple allow lists, so one can add to a standard list in a project (after review). And also deny, so one can remove a few without exiting completely from common lists.
Google is also in a great position wrt distribution - to get users at scale, and attach to pre-existing revenue streams. Via Android, Gmail, Docs, Search - they have a lot of reach. YouTube as well, though fit there is maybe less obvious.
Combined with the two factors you mention, and the size of their warchest - they are really excellently positioned.
I agree with you. There is also a move toward "agents", where the AI can make decisions and take actions for you. It is very early days for that, but it looks ike it might come sooner than I had though.
That opens up even more potential for influence on financial decisions (which is what adtech wants) - it could choose which things to buy for a given "need".
Pivoting every month! After 24 month and after raising VC money. That does not sound good. There are very few things that can be uncovered/learned in one month that would justify pivoting away from it, even in B2C. Even while throwing things at the wall and seeing what sticks, one needs to actually wait for the thing to actually hit the wall, at the very least...
Democratic governments tends to create and maintain a lot of checks and balanced on corporations. Things like workers rights, consumer rights, product quality, environment protections, etc etc. Many things that corporations would love to do to make more money are illegal. That makes it harder to earn money. Removing such barriers seems like a pretty straightforward motive to me.
One would never reach zero. And it would be challenging both to define and police laws against advertising. But to get to a world with drastically less advertisements than today seems doable.
I have also found this kind of approach very valuable. With defensive code, where the code itself does a bit of the kind of correctness checking that one might conventionally put into tests, it works even better. Pre/post- conditions in functions for example.
But the TV functionality does not have to work if one does that?
We should really have laws that makes it illegal to not function without connectivity.
Not the team, just follow ARC on-and-off as a ML engineer. I think it will take a few years (at least) to see the impact of ARC, especially the more conceptual works. Those are more close to basic research than applied - It will take time before the lessons are transferred to applications (that also requires considerable R&D).
But more importantly, current LLM-based systems and the in-the-spirit-of-ARC-systems have quite different goals. The ARC challenge is intended to measure and build system which can learn efficiently - that is, be able to solve a novel task with very little new data. Ref F. Chollet paper "On the Measure of Intelligence".
Current LLMs do not care for learning efficiency at all - actually the strategy is completely opposite - they aim to utilize ss much data and compute as possible to make the most capable system (at least on task that are somehow spanned by the training data). Which works well, but is for sure quite costly and it might also limit applications to those that not require a lot of learning at runtime (we still do not know how far we can take in-context learning).
ARC brings in a fresh perspective, but I expect it to take several years for the approaches to really start cross-pollinating.
Not Greg/team, so unrelated opinion. o3 solution for ARC v1 was incredibly expensive. Some good ideas are at least needed to take that cost down by a factor 100-10000x.
The task you mention require intelligence but also a robot body with a lot of physical dexterity suited to a designed-for-humanoids world. That seems like an additional requirement on top of intelligence? Maybe we do not want an AGI definition to include that?
There are humans who cannot perform these tasks, at least without assistive/adapted systems such as a wheelchair and accessible bus.
Will have to look closer at picoruby. I do love MicroPython development. One support thing is how easy it is to combine with C, because to get the most out of microcontroller devices one really needs to use low-level languages from time to time.
We have built a rather large toolset for detecting and treating a range of cancer variants over the last decades. With meaningful changes in expected outcomes. Considerable drop in mortality, for example. Far from cured though. Alzheimer's might be much trickier. ..
Being _perceived_ as having the best LLM/chatbot is a billion dollar game now. And it is an ongoing race, at breakneck speeds. These companies are likely gaming the metrics in any and all ways that they can.
Of course there are probably many working on genuine improvements also. And at the frontier it can be very difficult to separate "hack" from "better generalized performance". But that is much harder, so might be the minority in terms of practical impact already.
It is a big problem for researchers at least that we/they do know what is in the training data and how that process works. Figuring out if there are (for example) data leaks or overeager preference tuning, that caused performance to get better for a given task is extremely difficult with these giganormous black boxes.
Denying oneself something nice/good (but unrelated). Or doing something that provokes a punishing response from someone/something else. Like picking a fight, or "failing" to pay bills, gambling.
Neither will help a victim of course, but self-sabotage never help (almost by definition).
Why would a startup do that? Hardware is low-margin difficult, capital intensive, long sales cycles, long rampups until revenue, volume sales are very trust-based (hard to get in when not established), expensive recalls, product compliance, etc. In addition the trend for the last 30 years of electronics is integration, completely opposite of modularization.
I am very pro open source hardware and software (and contribute to both) - but I find it completely understandable that there are few startups/companies in this space.
I think that open-source modules are valuable even without open-source ICs. At least for may classes of genetic IC/modules there are many vendors of an IC, so one could have multi-source for the IC, to reduce reliance on a single vendor/source.
It is becoming more and more feasible to do things software-defined, with a generic microcontroller being the core IC. I would love to see some projects for DC/DC power converters for example around a microcontroller with open software. This is a bit tricky because it needs to be safe, but in theory doable. Also need to be EMC compatible, which will require a lot of testing.
Generally going beyond hobbyist requires much better QA than what most current open source hardware projects are willing/capable to do.
Environmental testing, EMC testing, regulatory compliance, etc. Some exceptions might be vendors like Olimex.
You describe a plan for building it. That is but one piece of the equation. How will you sell it? Why should people pick you over established AWS/Azure/GC? Will you target new projects, or get people to migrate? How does the strategies there differ? What is the timeline to becoming profitable? How will you fund it in the meantime?