I used General Assembly to learn frontend web development about 5 years ago. It was a positive experience, the two types of people I saw succeed were
1. people with some kind of tech background already and
2. people who were really smart but not technically oriented (we had 4 former professional artists and musicians in my cohort, one of whom was a Princeton undergrad).
The people who failed were people who had no curiosity to genuinely grow their knowledge and were just there because they wanted to check a box before getting a better paying job.
I've found that my curiosity drops commensurate with my savings (and therefore the chance I'm going to get kicked out of my home, my belongings repossessed, and find food unaffordable). If people are so insecure that this opportunity to retrain an entire workforce was squandered - the money was there, the desire was there, the expertise was there - it comes off to me as more of a, "If you owe the bank $1 billion...", sort of problem, for society.
I'm not OP, but George Hotz said in his lex friedman podcast a while back that it was an MoE of 8 250B. subtract out duplication of attention nodes, and you get something right around 1.8T
When it comes to layoffs, no one wants to be the first company to do it, and no one wants to be the last company to do it. Once a few of the industry leaders started doing it, it gave others top-cover to start the process. At this point they need to do the layoffs so if they report bad sales numbers for the quarter they can say "we saw this coming, thats why we did some layoffs alredy. we should be good now". And if it turns out the didn't "need" to do it, they can just hire the people other companies laid off.
This is how I see it as well. Many might have wanted to do it, but they didn't want to be the first. As soon as the big ones started doing, even just one or two, the floodgates opened and now everyone can blame it on "everyone else doing it", "it's the environment", "the economic reality", etc.
I'll take the other side of this argument and point out: it probably took decades if not centuries to start using paper currency. Will crypto fail? Maybe. But to say that people building an alternate currency to the one we currently have are "scumbag" is a bit much. They are building a financial system that we might one day decide is better than what we currently have available to us. Sure, some are opportunistic bad-actors, but 99%? I don't think thats fair.
The first people to accept paper were almost certainly scammed. That's because the thing was totally unregulated in its first hundreds of years of use.
MY wife and I have homeschooled our two kids through high school. It's been a positive experience for everyone. MY response to concerns that you might not have the skills, I bet whatever skills you lack can be more than made up for by teaching at a ratio of 1:1 and being able to tailor things. If you need help my wife has a homeschool blog with a lot of resources to help you get going: organizedhomeschooler.com
Couple things stood out:
13.4% of companies see themselves as leaders in AI. I think people are probably being a bit ambitious with how much of a "leader" they are.
79% are using some form of OpenAI GPT. I get that that is the easiest entry point, but are companies really that cavalier about putting their data and their customers data on someone else's hardware? I get the cloud use case, but this sounds more trusting than I would be with emerging technology.
only 26% are customizing base models in any way. To me that would be table-stakes for using LLMs in any way.
I guess the main takeaway is just how early we are in this movement, and how much work there is still to do.
100% agree, I think the 26% will greatly increase over time... or the ones that don't will decline as a business over time.
the 13.4% is likely leaders in ML for some specific use case like fraud or recommendations. it would be great to have access to raw data with anonymized demographics.
start-up idea if anyone is bored: disrupt car shipping. When I moved from Seattle to Raleigh shipping my car was BY FAR the biggest headache. It's a 20th century business where you have to call people, who will bid for you, and you have no idea if your bid is accepted. Agents will promise a low price to get your business, then tell you a week later that bid was too low, you need to double it. I eventually just gave up and drove it cross country.
Businesses like these are hard to "disrupt" because despite all the fancy websites and apps, pricing algorithms, scheduling, reminders, live video feeds and whatever else you can build, underneath it all the success or failure still depends on some guy showing up at your house, picking up your car and trucking it across the country. Unless you can automate that part nothing will fundamentally change.
It goes deeper than that. Even after all the sophisticated systems and automation, a huge amount of logistics is still email-driven, and it seems next to impossible to get account managers, carriers and customers past that.
I think the expectation is even worse for the car case than anything since so few people bother shipping cars to begin with. But just looking at moving and shipping, even then no one has really beaten the model that Uhaul has created. Have the 7/11 or big box hardware store parking lot and clerk be your brick and mortar and labor force. Have a very basic website match you to pickup and dropoff locations. Buy a bunch of cheap aluminum trailers that won't rust out on you. They will even sell you overpriced cardboard boxes they probably get a ridiculous markup on. I don't think you can lighten this sort of business much more than that, and yet the profit margins still aren't as high as you'd expect from such a skeleton crew of a business. 16% last earnings.
"Disrupting" this is quite a challenge due to the nature of the service and the logistics around this type of freight. You either pay for someone to drive it which will be on an ad hoc basis when someone is available and interested or you pay to ship it on a Class 8 hauled flatbed or car trailer.
A possible brokerage opportunity would be to inquire with Carvana or Carmax if you can buy slots on their cross country shipments as capacity allows, and sell those to customers. You'll be at the mercy of their schedules and capacity, but could be a service tier along with other more "white glove" services (dedicated driver "hot shot", semi shipment, etc). You could also buy capacity on freight rail cars shipping back empty to factories (depending on destination and routing). High touch biz, beware, there be dragons.
I've dealt with this, I've had friends deal with this, etc. Every person I've known who shipped a car was screwed over by it. Even military people. You're lucky they told you up front that your cost was double. I had a shipper tell me that I they were doubling the fee about 45 minutes before they delivered it (I called their bluff and said I'd just report it stolen - it was an $800 rat rod Miata though).
I think the fact of the matter is, shipping cars is not going to be cost effective for an individual. Trucks are expensive, putting fuel in them is expensive, paying a trained person to drive one is expensive.
Maybe a company like Carvana could offer this as a side service, drop your car off at a local Carvana, and eventually pick it up at another location. Even then, I think the costs are going to be order of thousands of dollars.
If one searches Reddit, there are brokers who will pass on wholesale car shipping prices which are lower than the prices of those one may find on Google.
Last time I moved a car halfway across the U.S. I just went online, submitted a request, got a quote, paid and it was all done. It was about as friction-free as something like that can be.
Just need to use a direct carrier like Intercity, NOT a broker. Brokers have infested Google SEO.
> Just need to use a direct carrier like Intercity, NOT a broker.
So because you mentioned it I just called them for transporting by enclosed carrier on the east coast between 2 cities. High end car enclosed carrier.
Intercity wanted $1995 for the same transport that the car dealer (a high end car dealer well established no less) want $1600 dollars for.
Intercity essentially said 'busy time of year etc etc'. Didn't give remarkable reasons why they would be better than anyone else. Picked up the phone right away and was otherwise business like.
Also their website to get a quote they want an 'auction' number. That made me call them (which I would normally never do.)
Additionally their price was higher than other brokers I had checked including the one that just moved a car for me (same route) not enclosed who quoted both enclosed and not enclosed.
Would like to point out that the job of a broker is not to match you with anyone but match you with people who they know and trust and have used. And specifically the broker I used (for the last open move) made a total of about $250 the rest paid to the driver on delivery.
Their point is that brokers are generally terrible. That you found an ok one is cool. We went through a broker for moving our possessions from California to Montana. Worst professional experience of our lives. They kept our shit at some lot for around six weeks while they tried to job out to other freight providers because the first one didn't work out. Or the next. Or the next. Finally they got someone who should have been on medical disability to drive our stuff up after the broker declared bankruptcy. Not sure how that worked exactly. I had to threaten to drive down there to pick up my own shit and load it into a uhaul or two.
Don't have experience with trans-continental vehicle transport but as part of an old job one task I had was "drive this truck from Utah to North Carolina.. work there for a few weeks then drive this van back to Utah". Certainly cheaper than designated shippers for a company...
As a consumer I'd assume there are hotshot truckers/services that would be used for this? People have mentioned similar "agent oriented" experiences with moving services and how it's opaque and unreliable.
'Dealer trades' are pretty common now. If you want a very specific car it might be the only way to get it these days since the used market is so anemic.
My local dealer somehow got a lead that I want a very specific year and trim GTI and twice now they have offered to bring one in for me.
sometimes, but they mostly get put on a transporter like any new car. If it’s something valid able (claaaic car, race car, etc) they’ll usually sprint for an enclosed hauler, which is am several times more expensive
This is the actual problem, because transporter trucks are almost always just used to bring a group of vehicles to a dealership from a rail depot.
The ones carrying cars across the country for moves are quite rare, unscheduled, and harder to find. They also can't carry anything else besides cars on the transporters - you might almost have better luck driving your car into the back of a box semi and letting them load other LTL loads behind it.
Did you ship your furnishings? I have shipped my car by simply including it in the move and it traveled with my other goods in the same truck. Not that it was a huge SUV.
Not sure I understand - is this direct to consumer play? If so, you will have no return customers because on avg people buy a car like once every 7 years - bad business
I shipped a car across the country a decade ago, and I didn't go through any of this. I used a website, it gave me a price, I paid it, and a truck showed up.
I rented a flatbed car trailer and attached it to the back of the moving truck. Even if you hire a moving company, the same should be able to be done, right? I've never used a moving company. Is this not something they handle? I at least had experience pulling a trailer, so it maybe wasn't as daunting of an idea for me.??
A large, established moving company can. Last time I moved, they just drove my truck into the semi-trailer along with the rest of my stuff then took it across country.
Yeah maybe do it via some kind of car rental. Someone going from Seattle to Minneapolis pays peanuts and insurance to drive your car. Someone else going from Minneapolis to Memphis after that. And so on.
Pretty much, yeah. And from my personal experience, uship is useless. Lots of independent people that charge too much due to small economies of scale (one guy with a goose-neck truck vs. a car carrier). The few times I've had to ship a vehicle cross country I've wasted time on uship before going back to a traditional broker and letting them handle it for less money and less hassle. It's still annoying dealing with the salespeople, but if you're good at playing that game you can at least use the usual negotiation tactics to get an even lower price.
Is it possible in the 21st century to have a city where artists all know to gather and make art together? I guess today the closest thing would be college towns? Or perhaps these cities still exist and I'm just not aware of them.
Berlin from about 1995 to 2015 would probably have been that place. Closest thing today might be Marfa, Texas, which obviously is more of a town than a city.
Definitely, but there are a good many people in general that probably consider a small rural town in the middle of nowhere to be a nonstarter in the first place.
I love Marfa, and have visited multiple times. That said, it's definitely a big sacrifice for someone to take a residency there because it's the type of small rural town that doesn't even have a 24 hour pharmacy or proper emergency care without a hour+ drive to a larger nearby town. It's definitely on the fringes of what's acceptable, so much so that it's genuinely surprising when you visit how much of an art enclave it is.
Marfa was able to draw many creatives in part because foundations (like the Lannan Foundation for writers) paid well for them to spend a season or a year there.
LA. You want an editing bay? Its for rent by the hour. Photography studio? For rent by the hour. Recording studio? For rent by the hour. Stage? For rent by the hour. There are creatives in every field here, a critical mass level that allows for an entire service economy tailored towards them to not only exist but grow.
Not knowing much about the topic I feel like your comment is more profound than other answers. Access to collaborators, tools and technicians is a real bottleneck that LA seemingly solves. The city is certainly geared towards producing media, if not art.
> in warehouses that had once served the maritime industry
loft apartments today were artist residences a generation ago, which themselves were just converted factories
when "work from office" became a thing and people stopped going to factories to work, the factories had to be converted since otherwise they would become worthless
now with "work from home", i think the previous generation "office" will be converted into artist residences again (plenty of windows and light, not too many legal hurdles)
get in now before offices go full circle into luxury "cubicle units"!
This is really helpful for me, the right level of example. I find most D3 examples to either be too simple or too complex, this nails the sweetspot of visually impressive, while still being practical.
What does "fine tuning" mean in this context? Does it mean you fine-tuned it on a specific code repository, or collection of code repositories and then had it do work in those repositories?
Broadly finetuning is any post pretraining training. Most of the time it is used in the context of fitting a more narrow task. In our case, it was the same training objective as the pretraining but meant to be more representative of what Replit users like to code. However, we were surprised by how well it boosted overall performance. Best guess: it's a) novel data and b) the model could take even more training!!
How feasible and effective would it be to fine-tune a model against an organization's private source code, resulting in an "internal" model that knows how to work with that org's stuff?
Could you, say, fine-tune the model every week with the latest merges? Every hour?
Finetuning is a relatively quick process. Training the base model is the expensive part (can take weeks and huge amounts of compute), whereas finetuning usually is only on the last few layers and can be done with much less resources. You could definitely have a "nightly" finetune model that is retrained every day or so.
Interesting - how would that work for a company that wanted to run their own codex model, on-prem, trained on their own code? Perhaps also trained on their dependencies?
Finetuning a smaller model leading to better performance seems like a significant finding that'll lead to a lot of companies fine-tuning their own internal "ChatGPT"s
You seem to know your stuff some, so I'll ask you a question on this: Are there any good books on all the different approaches in this space, or is it all too new and fast moving for such a thing?
There are no books on Large LMs but almost any resource about neural networks covers fine tuning. I like the FastAI courses, and these do cover language models.
When you fine-tune it, do you train just the head/last few layers or do you also unfreeze the model afterwards and retrain the whole model with a very small LR for a few epochs?
You can take a network and its weights that someone else trained, and use that pretrained network to train on your own data, which is likely to be a better starting point than random weights.
1. people with some kind of tech background already and
2. people who were really smart but not technically oriented (we had 4 former professional artists and musicians in my cohort, one of whom was a Princeton undergrad).
The people who failed were people who had no curiosity to genuinely grow their knowledge and were just there because they wanted to check a box before getting a better paying job.