fallingmeat

haha well they did get that cool nickname

This is why you should always synchronize you cloud with your other cloud, and the NAS in your closet, and floss after every meal, and take daily vitamins.

this is so much cheaper than fine-tuned model; it would make sense to try and use the multi-shot prompt in the 3.5turbo model. plus, there's the hundreds/thousands of training items that you wouldn't need to create....fml

I’ve tried to identify parts of the prompt in the output, hopefully recognizing someone trying to dump the prompt, but that’s pretty tough to do. I’m sure even fine tuned models are susceptible to spilling proprietary training data

i think that would be an even more interest post.

im also curious

not bad! eager to see how the economics work out. is openAi endpoint cost scalable for a "free" or "free-ish" service like this (seems like there will be a very liberal use of queries)?

oh it's nondeterministic! yeah this is fun

Do your part- make a review and put a dent in

isn't this what mbse (i.e. represented in sysml) is for?

whoa! this is amazing. nice work my friend

Wi-Fi jammer?

Location: Atlanta, GA

Remote: Yes or commute (~quarterly)

Willing to relocate: No

Technologies: Engineering management; Avionics embedded systems engineering (RTCA/DO-178C); Autonomous systems; Formal methods

Resume: https://www.linkedin.com/in/thenatefisher

^ this guy knows what he is talking about :). sup man

They do have a flight test program.

what are the killer apps for this? maybe someone more bullish can summarize who this customer is?

shouldn’t matter who does it or for how much, as long as the sw lifecycle process is followed (which would include supplier oversight). Actually writing code, btw, is the cheap/easy part- designing correct system and sw requirements, along with verification, is much more expensive.

+1

To anyone who has experience with Arc24; How does this compare to the MBSE/SysML framework from OMG? https://www.omgsysml.org/what-is-sysml.htm

Have you explored using a formal verification tool to prove out the safety/liveness properties of your (or the end user's) autonomy specifications? Could you model the system in something like Spin/NuSMV/PRISM and then ensure that certain properties hold (useful states are reachable and dangerous ones are not)?

(1) Model checking first (using a tool most appropriate for the domain): UPPAAL, SPIN, NuSMV, Z3, TLA+

(2) Model based design (less time spent on less value-added concerns...plus, capturing even more benefit from (1))

Does this use tactics to identify contradictions, tautologies, subsumptions, etc in conditions? How does this work?

Formal methods has a lot of practical application but the tools and techniques are very inaccessible to the average SME. We need better tools. For example, here is a demo of how to use an SMT solver to write better system requirements: http://slra1.baselines.cloud/

Good morning all, creator here- I made this demo for you to explore a requirements engineering technique. Defects in the requirements engineering process have outsized cost in the overall product development cycle, especially in safety critical applications. Adding just a dash of formal methods in authoring tools can catch defects as they are introduced and provide timely feedback to engineers. This is a quick example of how to use a context-free grammar and an SMT solver (Z3 in this case) to automatically find logical contradictions, redundant conditions and inconsistency of entity names. Enjoy!

Another use case: verifying natural language requirements. As an example: http://slra1.baselines.cloud

Anyone know if Activity diagrams can be made in this way? (Note activity diagrams are not exactly flow charts, they have some rules)

well said

They both support discoverability in different ways

"...they want you to look up so they can look down on you!"

Best line from that movie

Another "popular" option is Yices: https://yices.csl.sri.com