It's an ironic situation; logically what should be the moat are the models, costing hundreds of millions of investment cost to train and operate so it would make sense if we see different provider focusing in different directions.
But right now we have 3-5 top contenders that are so evenly matched that the de-facto sticking point is mostly the harness, ie. the collection of proven plugins/commands/tools/agent features that are tuned to the users personal workflow.
That's why the frontier LLM companies are now spending a lot more to license exclusive proprietary training data from private sources in order to gain a quality edge in certain business domains.
But those holding said proprietary data have figured out they’re holding the cards now and have gotten a lot smarter recently. Companies are being very careful about what gets used for inference vs what they allow to be used for training.
I don’t see the core models getting dramatically better from where they are now. We’ve clearly hit a plateau.
Really? I mean I see regularly as I'm coding how much better it could be simply by running obvious prompts for me.
When I use the planning mode and then code the success rate is much higher. When I ask it to work on specific isolated chunks of code with clear success/failure modes the success rate is again much higher.
Now imagine a world where it recognizes that from my simple throw away non specific prompt. If it was able to fire off 20 different prompts in quick succession it could easily cut my time spent in front of the screen by a third.
The patterns are obvious but they don't do that right now because it's a lot of compute.
We'll be looking at this time where there's a progress bar showing context space the way we look at the Turbo button.
Because the truth is to get the baseline I'm talking about is a finite amount of compute at a certain point.
That sounds like spin to me. If there were a clear "quality edge" in "certain business domains" stemming from "exclusive proprietary data", someone would have been exploiting it already using meat computers.
But no, businesses are dumb. They always have been. Existing businesses get disrupted by new ideas and new technology all the time. This very site is a temple to disruption!
Proprietary advantage is, 99.999% of the time, just structural advantage. You can't compete with Procter & Gamble because they already built their brands and factories and supply chains and you'd have to do all that from scratch while selling cheaper products as upstart value options. And there's not enough money in consumer junk to make that worth it.
But if you did have funding and wanted to beat them on first principles? Would you really start by training an LLM on what they're already doing? No, you'd throw money at a bunch of hackers from YC. Duh.
Frontier labs are paying the same constellation of firms offering proprietary data and access to experts in their fields to train LLMs.
They are neck-and-neck only because they are participating in the arms race. The only other way to keep up is mass-distillation, which could prove to be fragile (so far it seems to be sustainable).
Meh. I think there's basically no benefit shown so far to careful curation. That's where we've been in machine learning for three decades, after all. Also recognize that the Great Leap Forward of LLMs was when they got big enough to abandon that strategy and just slurp in the Library of All The Junk.
I think one needs to at least recognize the possibility that... there just isn't any more data for training. We've done it all. The models we have today have already distilled all of the output of human cleverness throughout history. If there's more data to be had, we need to make it the hard way.
Ok, maybe pretraining is now complete and solved. Next up: post-training, reinforcement learning, engineering RL environments for realistic problem solving, recording data online during use, then offline simulation of how it could have gone better and faster, distilling that into the next model etc. etc. There's still decades worth of progress to be made this way.
" There's still decades worth of progress to be made this way."
That's not true. Moreover the progress can slow to a crawl where it's barely noticeable. And in that world the humans continues to stay ahead - that's the magic of humans. To be aware of surroundings and adapt sufficiently whilst taking advantage of tools and leveraging them.
This is an interesting theoretical statement that does not survive a collision with reality. The long-tail expert RHLF training is effective. We have seen significant employment impact to call center employees. This does not mean its progress will be cheap or immediate.
The quality edge hasn't shown up yet. If this strategy actually works then the quality improvements will only become apparent in the next round of major LLM updates. There's a lot of valuable training data locked up behind corporate firewalls. But this is all somewhat speculative for now.
To stop this, I today put most of my Amazon Redshift research web-site behind a basic auth username/password wall.
It's all remains free, but you need to email me for a username and password.
If I put in time and effort to make content and OpenAI et al copy it and sell it through their LLM such that no one comes to me any more, then plainly it makes no sense for me to create that content; and then it would not exist for OpenAI to take, or for anyone else. We all lose.
CloudBolt’s survey also examined how respondents are migrating workloads off of VMware. Currently, 36 percent of participants said they migrated 1–24 percent of their environment off of VMware. Another 32 percent said that they have migrated 25–49 percent; 10 percent said that they’ve migrated 50–74 percent of workloads; and 2 percent have migrated 75 percent or more of workloads. Five percent of respondents said that they have not migrated from VMware at all.
Among migrated workloads, 72 percent moved to public cloud infrastructure as a service, followed by Microsoft’s Hyper-V/Azure stack (43 percent of respondents).
Overall, 86 percent of respondents “are actively reducing their VMware footprint,” CloudBolt’s report said.
It is easier to do in the cloud than it is to do with actual hardware though, because you'll need enough hardware to do the migration. There is a capital moat around that.
I feel like the company that can figure out how to 100% safely live migrate any VMWare workload to another "cheaper" solution, will do quite well.
Do you not know how ebay works? You put in the maximum price you're willing to pay, and if you win you're paying 2nd highest bid + 1. So you don't save any money by starting with a low bid.
From what I've seen discussed, it seems some percentage of "sniping" is to attempt to obtain both "winning bid" and "lowest possible price" (note, not the same as "max willing to pay for the same item"). The sniper is trying to hide interest, so as not to attract other interested bidders, and therefore grab "a great deal" of a small increment above the starting bid price.
And this probably appears to work enough times in the snipers favor to trick them into thinking it is a winning strategy, whereas they likely would have won the same auctions in the end by just bidding that 'minimum' as their maximum bid. But as they can't easily (i.e., without expense) A/B test their strategy, they get no feedback that sniping isn't really helping them like they think it is helping them.
> But as they can't easily (i.e., without expense) A/B test their strategy
There also isn't really any detriment. At worst, the sniper is making the same bid they would have made otherwise. If the opposing bidders are not purely rational, and have not put in their actual maximum bid, then sniping can deprive them of that opportunity and thus lowers the hammer price.
And bidders are not purely rational, especially when the items are not purely utilitarian. Getting notifications that you have been outbid has an emotional effect, as does having time to think about raising the bid.
they notify the bidder when they're outbid, and the incremental price increases can make it tempting for someone to adjust their idea of their max price. sniping deprives them of that opportunity.
Taxis and B&Bs and Hotels all still exist and compete with Uber and AirBnB.
Even at the same price there are valid reasons why many people prefer an Uber over a Taxi, in particular the predictable pricing and globally consistent UI.
Uber changes the pricing depending on the demmand. Taxis can't do that, they are regulated by law in most of the west.
Predictable my ass. You have been lied to.
And btw all over the world you rise up your arm, and the taxi stops, I think that is a pretty consistent user interface that anybody in the world can understand. I have to help my aunt each time she needs an Uber.
Please don't pretend to be dumb, "predictable" in this context means that you know the final price of the trip before you start the journey, instead of having an odometer.
And randomly hoping for a taxi to drive by maybe works as long as you exclusively travel between airports, train stations, and downtowns of major cities, but if you're even slightly more remote than that you'll have to call or use some random app to get the Taxi to pick you up.
If the problem starts to become big enough, I'd expect airsoft venues to offer special streaming or non-streaming times, depending on which group is bigger. Similar to how Saunas offer special clothed or women-only days.
H200 rental prices currently start at $2.35 per hour, or $1700 per month. Even if you just rent for 4h a day, the $200 subscription is still quite a bit cheaper. And I'm not even sure that the highest-quality open models run on a single H200.
I'm pretty sure OP wasn't talking about the management hierarchy, but "from the top" in the sense that it was big established companies inventing the cloud and innovating and pushing in the space, not small startups.
That could be, I was definitely thinking of management hierarchy since that difference has been so striking with AI.
A lot of my awareness started in the academic HPC world which was a bit ahead in needing high capacity of generic resources but it felt like this came from the edges rather than the major IT giants. Companies like IBM, Microsoft, or HP weren’t doing it, and some companies like Oracle or Cisco appeared to thought that infrastructure complexity was part of their lock on enterprise IT departments since places with complex hand run books weren’t quick to switch vendors.
Amazon at the time wasn’t seen as a big tech company - they were where you bought CDs – and companies like Joyent or Rackspace had a lot of mindshare as well before AWS started offering virtual compute in 2006. One big factor in all of this was that x86 virtualization wasn’t cheap until the mid-to-late 2000s so a lot of people weren’t willing to pay high virtualization costs, but without that you’re talking services like Bingodisk or S3 rather than companies migrating compute loads.
Sure Amazon was a big established co at the dawn of the cloud, and a little bit of an unexpected dark horse. None of the managed hosting providers saw Amazon coming. Also ran's like Rackspace and the like where also pretty established by that point.
But there was also cool stuff happening at smaller places like Joyent, Heroku, Slicehost, Linode, Backblaze, iron.io, etc.
For me, I actually feel much more motivated to write good and accurate documentation knowing there will be at least one reader who is going to look at it very closely and will attempt to synthesize useful information from it.
Same with my old open-source projects, it's kinda cool knowing that all the old stuff that nobody would have ever looked at anymore is now part of a humanity-wide corpus of useful knowledge on how to do X with language Y.
That's why it's just "mostly" false, but 'empty' is a word with a specific meaning, and claim here was that the port is literally empty of ships. (or, in the case of the Twitter message they show, that there's only one single ship in the harbor)
But right now we have 3-5 top contenders that are so evenly matched that the de-facto sticking point is mostly the harness, ie. the collection of proven plugins/commands/tools/agent features that are tuned to the users personal workflow.
reply