Working on an entry and trying to decide how heavily to use Gemini given that if the app scaled to high usage it may end up being expensive compute. In the real world you would balance how much you use AI but for the contest do you think it makes more sense to let loose?
Welcome to the forums!
The following is my personal opinion, but is also the guideline that I’m using as I’m developing for the contest. So take that as you will.
My view is that we should be demonstrating how Gemini can bring true value to the application. Not how it is being used for trivial things, or things that can be better done using other means.
There are several judging criteria that I think apply:
Is the submission surprising to those that are well-versed in Large Language Models (“LLM”)? (maximum 5 points)
I read this as “surprising in a good way” not “surprising that nobody should even think of doing that”.
Is the submission implemented through the use of creative problem-solving approaches? (maximum 5 points)
Throwing Gemini at every problem is not “creative”. Using it as a vital tool to do things that could never be done before is.
Is the solution well-designed and adhere to software engineering practices? (maximum 5 points)
There are many components to this, but one is to not use a chainsaw when a scalpel is more appropriate. Use the right tools for the job. Is Gemini the right tool? Maybe - maybe not.
However you decide, make sure you make it clear in your presentation why you chose that tool. Make the judges understand that Gemini was chosen because it is the right tool - not just because it was the tool that was available.
Good luck in the competition!
2 Likes
Really awesome thoughts thanks! I’m not however sure how it answered the question at all. I’ll try word is better…
If faced with decisions in choosing an execution that are really powerful but expensive compute vs. being mindful of making it more efficient and scalable, which path would you choose?
The answer is obvious for a production environment but for the contest you can push the tech harder and showcase more interesting things. Some could say “screw production the cost will come down over time” for example.
My answer stands. {:
Whatever your decision - make sure you justify it as an engineering decision as you explain it to the judges, considering the above criteria. (See my third point about software engineering practices.)
3 Likes
very great comment sir. thanks
I’ve won hackathons before by just tossing a maxed out GPT-4 at the problem. We justified it by comparing the AI costs to the cost of getting an engineer to do it. Don’t pitch it as, “ummm this is kinda expensive,” pitch it as “This frees up engineering time to work on something more productive!” Round up where you can on cost; it should seem logical enough that the judges won’t have to pull out a calculator.
It’s very subjective. Would it be done only once in reality? Is it super intensive like AI agents? Or are you doing something ridiculous like selling a dollar of compute for half a dollar?
AI agents are probably the worst example I can think of, especially those that work in a team, but compared to the cost of a human, you might be able to justify it. Even if it were a game, people spend lots of money on games. Hackathons are a good place to play with ideas that you wouldn’t do in reality anyway.
1 Like