Searching...
19 results for “machine logic”
...to put logic on the outside to orchestrate. And it's not just a matter of routing between, you know, this LLM or that LLM. Sometimes you run code. Sometimes you use a tool. You know, you'll access the database. You'll call, you know, a weather API, w
4:32 / 5:34
from transcript4:32 – 5:34
But when it comes to mission critical applications, if you're doing something in accounting, if you're doing customer support, obviously if you're a developer and you're pushing code to a server, it needs to be a lot more reliable. And and that's why
3:46 / 4:50
from transcript3:46 – 4:50
“Why traditional software only automates things that are easy to explain”
...In fact, automation, machine tools, sewing machines, typewriters, adding machines Right. Things that are easy to explain to a computer. There may be things that are very hard for people to to to do, b
18:05 / 18:30
framework18:05 – 18:30
AI side of things and go back into the more deterministic, perhaps even human led part of work? Is there is there a trigger that comes into play? Is it just a complexity point? It's basically a human in the loop connector. Right? How our AI workflows
35:12 / 36:22
from transcript35:12 – 36:22
...ship the machine out to a different manufacturer. You can do that on prem within a couple of days rather than waiting for years to happen. We also have like a marketplace of skill set. So let's say you're a developer wanting to build a skill set for
6:46 / 7:48
from transcript6:46 – 7:48
handles your entire digital life and is way smarter than everybody. It's like it's operating in a and so it's an interesting leap of faith to go from cloud code becomes that, which it, like, in some ways is there's some avenues for that, but I do thi
3:42:00 / 3:43:01
from transcript3:42:00 – 3:43:01
Adprobic makes great products. PlotCore is fantastic. Co worker is fantastic. But they are cranes of silicon doing matrix multiplication. They don't have consciousness. They don't have an inner monologue. You take an alum and train it on pre 19 '16 o
0:00 / 1:03
from transcript0:00 – 1:03
...the business logic that was produced by the coding agent. You know, I think a lot of people are seeing this that the more declarative and clean and dry the code that you're writing and the more type safe the code that you're writing is, the more you
18:44 / 19:47
from transcript18:44 – 19:47
It's like a caching of the generated code that I can then, like, incur any inference time cost. It's just the actual code at that point. Yeah. I invested in a company called E2B, which does code sandbox, and they powered the LM arena web arena. Yep.
47:15 / 48:17
from transcript47:15 – 48:17
...machine tools, sewing machines, typewriters, adding machines Right. Things that are easy to explain to a computer. There may be things that are very hard for people to to to do, but they're easy to explain. So it's hard for you to drill a hole a 100
17:12 / 18:30
from transcript17:12 – 18:30
...logic, right? Where I'm basically, taking context that I retrieve somehow from databases. I assemble that into a prompt. I run the prompt. And then I occasionally invoke tools. Maybe I do that with MCP or something like that with an external server.
26:15 / 27:21
from transcript26:15 – 27:21
symbolic systems, I think, or like the people that think that discrete reasonings or F statements and knowledge bases, whatever, this is the way to go. And so there was there was a merging of these two worlds where the way AlphaGo worked is it had a
24:00 / 25:30
from transcript24:00 – 25:30
It worked. He had no idea why. So he set out to build a mathematical model of how LLMs actually function. The result? A series of papers showing that transformers update their predictions in a precise, mathematically predictable way. In controlled ex
0:44 / 1:45
from transcript0:44 – 1:45
you can use this in the wrong way. And the example I used earlier was if, you know, you buy Jira for example, but you use it for your shopping list, Atlassian is happy about this but that's not what they built a product for. Right? But you can use it
33:47 / 34:50
from transcript33:47 – 34:50
the way we have learned multiplication tables, now you know exactly what to do next step? Right? You write 769 and then one zero two five, and then you know exactly. So at each state of that process, your prediction entropy is very low. You know exac
8:59 / 10:01
from transcript8:59 – 10:01
is a great threshold, but it feels a little synthetic. We're always evolving. Right? So we're we're always learning more here. We we have some built in evals that allow you to kinda test, you know, in the product and and and to to to help the LMM to
36:00 / 37:03
from transcript36:00 – 37:03
They basically created an instruction tuning framework that teaches LLMs to do symbolic planning, which basically means that the LLMs think about step by step or chain of thought in a smarter way by making them generate explicit state action state ch
1:00:00 / 1:01:02
from transcript1:00:00 – 1:01:02
But the agent agents are doing things autonomously, doing things for minutes at a time, hours at a time, etcetera. Right? Reasoning is doing things for tens of seconds at a time. Right? And then coming back with an output that I still need to verify
4:33:45 / 4:34:45
from transcript4:33:45 – 4:34:45
concept to modern day AI. If you think about large language models and predicting what comes next, it's not like these large language models necessarily understand English. They're just really, really good at predicting states and the next state, I e
28:26 / 29:37
from transcript28:26 – 29:37
Have a podcast?
Run your podcast through Clypt
Get ranked clips, hooks, and ready-to-post copy from your own episodes. Free to try.