Meta has spent massive sums of money to train these models and they've released the models to the public. You can fine-tune the models. You can see the source code and the architecture of the model. The EULA is commercially-friendly.
You are free to quibble over how truly "open source" these models are, but I am very thankful that Meta has released them.
Then there's the question of how the electric grid will sustain this significant new increase in demand. As with mandates banning gas cars, the grid supply and stability is always regarded as a mere detail to be sorted out later.
Slack is great in small doses, but a complete time-waster beyond a certain threshold. I don't have ADHD, but I still have had problems with Slack. Don't feel bad about shutting all notifications off, or, better yet, closing the app entirely, and only opening it once or twice a day. You will almost instantly feel better. Some people may get annoyed that you're not instantly available, but if you're not able to get your work done then many more people will get annoyed. There is no perfect solution.
I have worked in both semiconductor manufacturing and drug discovery industries. Reliably and profitably producing 5 nm chips is an extreme engineering challenge, but- it is an engineering challenge. Drug discovery is a question of science and requires a fundamentally different mindset that semiconductor manufacturing. Human biology is much more complicated than manufacturing chips (and that is extremely complicated); drug discovery is about "unknown unknowns". Discovering a drug that has the intended effects without causing terrible adverse effects is something that some of the best-funded companies on the planet struggle with.
If there's one thing I learned with biomedical data modeling and machine learning, it's that "it's complicated". For biomedical scenarios, getting more data is often not simple at all. This is especially the case for rare diseases. For areas like drug discovery, getting a single new data point (for example, the effect of a drug candidate in human clinical settings) may require a huge expenditure of time and money. Biomedical results are often plagued with confounding variables, hidden and invisible, and simply adding in more data without detection and consideration of these bias sources can be disastrous. For example, measurements from lab #1 may show persistent errors not present in lab #2, and simply adding in more data blindly from lab #1 can make for worse models.
My conclusion is that you really need domain knowledge to know if you're fooling yourself with your great-looking modeling results. There's no simple statistical test to tell you if your data is acceptable or not.
One of the big differences between the US and Europe in this regard is the much larger proportion of bicyclists in Europe. In Europe, the larger number of bicyclists trains you as a driver to always be on the lookout for them. Driving past a bicyclist in the US is a much rarer event. Drivers should of course always be vigilant for anything in their way, but as a practical matter it's easier to mentally drift off when bicyclists are much less common.
With this kind of uncertainty, is it any wonder that many people don't trust medical science when we are told to behave in a certain way, or when we are told to take a certain medical treatment?
How many years have we been told that saturated fat is the nutritional devil? This is a great example of why censorship of "misinformation" is not justified- who is to say what constitutes misinformation?
There is PhD level math involved. And yet, ML (deep learning in particular) is much more of an empirical endeavor than many would like to admit. A deep understanding of the underlying mathematics does not necessarily give you a better model. Modern models are so complicated that no one can reason through them. Parameter spaces are non-convex and fully of ugly pathologies that make neat and tidy analysis methods useless.
From one perspective, it is disheartening that a deep understanding of the underlying methods doesn't necessarily win the day. From another, it is quite remarkable that having good implementation skills and a methodical mindset can get you quite far.
Calling YOLOv5 a "fraud" is a bit harsh. It has many excellent aspects for practitioners: easy to use, fast inference time, scalable model architecture, and it has many helpful utilities built-in for model deployment. In my experience, in real use-cases the models achieve about the same precision / recall / mAP as well as "state of the art" methods that report better stats on benchmarks.
What I'm hearing you say, loud and clear, is that you no longer have a sense of a purpose in life. Drugs (legal or otherwise) will not fill that void for you. I do not believe that you can rekindle a sense of purpose just by thinking about it. I would advise you to try new things that sound interesting to you, and keep exploring until you find something you can pour your heart and soul into. Probably something different from, or at least not directly related to, tech.
I share your frustration. If you're trying to be realistic and straight-forward about the limitations of AI/ML, you're getting little interest from investors. Here's an anecdote. I was at an "AI in drug discovery" conference 2 years ago. One of the presenters, the founder of a drug discovery start-up, emphatically made the claim in a talk that "we should deliberately overhype AI in drug discovery, to raise awareness of what we as an industry can accomplish". I was gobsmacked by this. And yet, in the social mixer afterwards, the investors I spoke with LIKED this approach- they said the boastful founder was bold and visionary, and they didn't care that he was exaggerating the capabilities of his company. So, that's why all we hear is hype- founders are just responding to investor incentives.
That's also why I work as a freelance consultant and not as a founder. I think that autonomous driving (rather, the lack of such) is going to be what triggers the next AI winter. Too much money and hype, too little results for too long- the rope is wearing quite thin from what I can see.
Your suspicion is largely correct, in my experience. I've worked as a consultant on a number of different AI / ML projects for start-ups. Most aren't doing anything all that new or groundbreaking from an ML point of view. Their real innovation is usually more about applying ML to industries / areas where it hasn't been used much before. But that doesn't get the investor dollars flowing in, so the founders try to make it seem like they have some radical new ML breakthrough. And in recent times, it has worked.
If there's one thing I've learned from the dotcom crash and the 2008 collapse, it's this: when the crap hits the fan, cash is king. Start-ups without profitable business models are extremely vulnerable. All of the pitches about exponential growth and glorious future profits carry little weight when investor psychology turns negative. So if you have a large pile of investor capital and you have investors who can't / won't claw their money back, you may be OK. But don't kid yourself that you are on an easier path.
As a computer vision practitioner, I would argue that SIFT is still very relevant today. In most real-world scenarios it seems to hold up as well or better than the deep learning approaches I've tried, and it is easier to implement and maintain. Failures are often easy to understand and mitigate. In practice I often end up using FAST or ORB features due to "good enough" accuracy but much faster processing rates, especially on embedded devices. Feature detection and matching is an area where "classical" computer vision is very much alive and well.
You're trying to act like everything is A-OK with this project, when it has in reality been a huge waste of public money. First, there has already been a large corruption scandal centered on this project, with people in prison because of associated corruption (search Alain Kaloyeros). Second, those Tesla employment requirement milestones have been repeatedly reduced to require fewer jobs, and the types of required jobs has also been watered down. The requirements in the original agreement are more stringent than the 1460 jobs required now.
I had high hopes for this project when it was announced. It is near my hometown, and I thought it might be a good boost for the local economy. All it seems to have been is a big siphon from public funds, but not a siphon to where it is needed.
The history of the Buffalo "gigafactory" is rather sordid... In a nutshell, Tesla and its subsidiaries made large employment and economic growth promises to the state of NY in exchange for massive subsidies (almost $1B total). The state wanted to stimulate growth in an economically depressed region, but those growth expectations have not materialized at all. And now, at least some of the NY state taxpayer financed equipment is apparently being shipped elsewhere or sold instead of helping the economy of Western NY. That's the story here. But the entire story goes much deeper.
Tesla fans don't like to admit it, but the Buffalo "gigafactory" project has been a textbook example of government waste on corporate welfare. Multiple people have gone to jail for corruption / bid rigging on the project. Tesla's subsidies were supposed to be contingent on achieving hiring and economic output goals, but those original requirements have been retroactively watered down multiple times so as to prevent subsidy clawbacks. A local journalist (Dan Telvock) has been detailing considerably more chicanery that has gone done at the Buffalo gigafactory.
This is what a speculative bubble looks like. Enjoy popcorn on the sidelines until people start pricing in financial reality. Manias sometimes take a while to fizzle out. Year over year revenue growth is actually flat to slightly negative for the past 2 quarters. If top-line revenue growth continues to stall out, that's often a catalyst for corrections.
I have found PyCharm to offer a good trade-off between data exploration and productionizing your code. It has the best Python debugger that I've used. You can also run Jupyter notebooks in PyCharm when that makes sense for you.
There's a reason why almost everyone pursuing autonomous driving uses lidar despite the high costs. It's because lidar has extremely good "recall" and "precision" in most driving conditions. In other words, if an object is present on or near the road, lidar will almost certainly detect it (high recall), and it has very rare false positive detections (high precision). Lidar also directly provides distance information. For pure camera based approaches, we've seen huge improvements in recall in the past decade, but unfortunately we still need orders of magnitude reductions in the false negative rate to be good enough for safety critical applications. I think pure computer vision approaches will get there someday, but we first need some fundamental changes in CV algorithms in my opinion. Hopefully with increasing adoption, lidar will get significantly cheaper.
Did you ever read about the obsessive determination with which Harry Markopolos worked to expose Bernie Madoff? He had a visceral obsession with Madoff, working for almost a decade to expose Madoff's scam. John Carreyrou likewise had a longterm fixation on exposing the Theranos scam. These weren't mere "disputes" over technicalities- they were all-consuming endeavors that became core to the identifies of the whistleblowers. They were most definitely "haters".
> There are of course some people who are genuine frauds. How can you distinguish between x calling y a fraud because x is a hater, and because y is a fraud? Look at neutral opinion. Actual frauds are usually pretty conspicuous. Thoughtful people are rarely taken in by them. So if there are some thoughtful people who like y, you can usually assume y is not a fraud.
Just look at recent history- Enron, WorldCom, Bernie Madoff, Elizabeth Holmes (and many more) were ALL widely celebrated by neutral third parties before being exposed as frauds. Neutral opinions are often neutral because they haven't done much research on a topic. Thus, they are often relatively uninformed opinions.
And these frauds were NOT conspicuous at all. They worked very hard to present the appearance of success. It took some dogged investigations from "haters" (by PG's definition) to reveal the truth.
To my knowledge, all he has ever said on the subject is: "I think people would be quite surprised if they knew how simple our methods are". You probably won't ever hear more information than that, until their strategies stop working.
What's your evidence that electric vehicle are more reliable than non-electrics? EV advocates make this claim quite frequently, yet I've never seen any evidence to support it. In the case of Tesla, their EVs are actually much less reliable than ICE and diesel vehicles (though I blame their problems mainly on Tesla's rush to production, and not on producing EVs per se).
You are correct- probability of success is (in my opinion) the single most important factor when evaluating options. Even if on average employees did well with stock options, the fact of the matter is that the distribution of payouts is very lopsided, not unlike professional athletes or actors. So most start-up employees will get nothing at all for their hard-earned options.
From personal experience, I have worked as an employee at 3 start-ups, and in only one of them did I get any payout at all from my options. And the amount of value I got from that one "success" was trivially small after multiple rounds of dilutive funding (financially, I would have been far better off flipping burgers in lieu of all that overtime). Most of my peers have had similar success rates with start-ups, with a few notable exceptions.
In fairness, Elon Musk himself has stated that Tesla has been several weeks from failure at multiple times in its history (most recently, in the fall of 2018). It's not like the critics were wrong about the financial health of the company.
I hear people say things like this about Tesla, but I am really skeptical. The company is still highly unprofitable despite having scaled production up significantly. It survives because of investor capital. That kind of strategy runs out of steam when investors stop believing the company's promises of future profits, or when a recession hits.