Are there platforms that make such training more streamlined? Say I have some definition of success for a given problem and it’s data how do I go about generating said RL model as fast and easily as possible?
We're working on an OSS industrial-grade version of this at TensorZero but there's a long way to go. I think the easiest out of the box solution today is probably OpenAI RFT but that's a partial solve with substantial vendor lock-in.
Typically speaking an LLM is the code driving the control flow and the MCP servers are kind of dumb API endpoints (find_flights, search_hotels, etc) say for a travel MCP.
With your product, how is the LLM made aware of the underlying data store in a more useful way than “func search(query)”?
It seems to be that if you could expose some precomputed API structure into the MCP for a given data store then the LLM could reason more effectively about the data rather than throwing search queries into the void and hoping for the best?
From what I have gathered their main differentiator is taking the approach of assigning each discrete data point its own "entity" definition that is independent but can be extended for each data provider.
So since its all represented by entities, you could treat them like any other vectorised data in your vector data store and use vector search.
It's a nice technique, but probably tricky if they ever venture into encapsulating endpoints in realtime for rapidly changing b2c applications (ratelimits/cronjob latency)
You should go. I was all set to go to Caltech or the Ivies and turned them down for mostly, though not entirely, financial reasons and went to a state school instead. Worked out mostly fine but much harder to get foot in the door and the opportunities and experiences you get in a place like MIT last a lifetime.
If you don't go, many doors will become closed or so much harder to get into that it's basically equivalent. Want to join a think tank, become a quant on Wallstreet, easily raise VC money at 23? These are all much easier with the credibility a MIT degree brings.
"Normal" degreed people essentially live a life where everybody assumes you are stupid until proven otherwise. MIT people are assumed to be smart until proven otherwise. That is a tremendous advantage in almost all contexts.
Not to mention the quality of your peers and the education itself. Don't pass up the opportunity to really challenge yourself and see what you can do. Sure, it's just undergrad and you aren't really solving anything of note but just being surrounded by the leaders of your field is inspiring and will push you to be your best.
Sample size 1 over here but especially MIT I would go. MIT is big and diverse enough that you can go as technical as you want, or study business, or…?
Anyway my 2 cents from a 30 something that has been around the block.
> Want to ... become a quant on Wallstreet...? These are all much easier with the credibility a MIT degree brings.
Do enough hard math EE at MIT and becoming a quant as your first job is a common initial career step. The money you'll make will then help you do whatever you want going forward.
Not sure how this would fit in with the OP's desire to get a CS Ph.D.; on the science track it's very easy to fall off, I'd expect CS to have more flexibility but I don't know.
Just wait until all software becomes a streamed H.265 video stream. That's certainly the future once latency allows for it. No piracy, infinite subscriptions, total control. But if you don't like it, don't buy it. That's life.
It's a different approach, coupled to automating the desktop office ecosystem, whereas we're coupled to web-apps, and web APIs alone.
Secondly, it is more complicated, a bit more like Leapwork, whereas we're targeting Zapier-level complexity.
Axiom is already too complex for many people (it's why we mainly target these no-code Zapier types). We've seen every marginal % increase in complexity reduces the number of people who can build bots significantly.
Essentially, each RPA product has chosen a power vs ease-of-use trade-off for different segments. We're fixated on the Zapier / Airtable people, not the traditional desktop RPA people, whom I think Microsoft are targetting.
I'm sure they will face the piper at some point, but you don't see people in Copenhagen living on the streets or in squalor. Probably because China makes all of their stuff and immigrants build all of their houses, but still.
Have you seen our housing stock, our housing supply, our infrastructure? There are plenty of jobs needing to be done. A virtually unlimited amount of progress is possible.
The problem inherently lies with the assumption that the free market will fix all and assuming that some entrepreneurial person is going to lead the charge and redevelop our country. What a joke.
The boom after the Great Depression that made the US into a superpower was driven entirely by government dollars.
My solution would be:
* Raise minimum wage to $30 an hour.
* Implement land value tax + higher capital gains taxes in order to stem inflation.
* Hire 10x the amount of construction workers, civil engineers, etc and provide them with the funding for projects.
Get to work. It's laughable that someone working at Snapchat can make 1M a year while the jobs that literally build our country are paying dogshit.
We need wealth redistribution but not via UBI, via actual development. Let's raze the slums and rebuild with shiny skyscrapers. It's possible. We just have to get away from capitalists getting increasingly rich while the rest suffer.
Also as an aside, my bet is that "programming" reverts to the mean within 10 years as a career. 20 years ago programmers were paid peanuts. Most of the heavy lifting of building software has already been done, AI, low-code, etc. is going to make it increasingly easier. Be ready.
All of this is GOOD, why does it cost 5 million dollars to build a site so that I can reserve a place in line at the DMV (made up example). The tech industry has enjoyed a monopoly on transmitting data. Building an app or a website should be as easy as opening an excel spreadsheet or a word doc. That is coming and your UI coding skills are going to be essentially useless very soon.
A single contractor will never be able to generate more money than a single software engineer. There is a reason software development pays so much.
I can go work for any company and help get rid of their employees or maximize efficiency. A simple program can save hundreds of man hours.
People who think building a website is cheap have no idea what they are talking about. Sure, a static site is easy. However, once you are building a site that needs to work with a system from the 1980s, you start to hit road blocks....and many many large business run on stuff from 80s. Places like costco, banks, etc.
