AI Caramba #12: Instant $$$
Custom GPTs are the future and they are EPIC at raising venture funding. Wait.
Me training my super powerful AI.
Hey. AI Caramba post? Yes. Who would have thought. Not me. XOXO.
Ok. My startup, AWDIO, is about to raise capital.
It’s a grind, but I love it.
Preparing, reaching out to venture people, not getting answers, getting answers, having calls, “can’t hear you, call back later”, “lets follow-up” (to never hear from them again) and so forth. Repeat. Cry. Repeat. Win.
Meet Turbo.vc
So I decided to try solving my challenge using AI, specifically GPTs. I’ve been told they can do anything and have replaced everyone in coding, marketing and copywriting so far.
What is a GPT, you ask? TLDR; It’s a chatbot that you can customise to whatever need you have and can solve complex problems → if you want to learn more
“But are there not other GPTs that do the same thing?” Yesno.
I found some that can read a deck or slide and comment really well. Not bad, but not exactly all of my challenges.
So I decided to build!
I put in 3 instructions on what to do and fired up my new best friendGPT, Turbo.vc :
Analyse a deck and come with suggestions for improvement
Identify 5 funds to talk to (based on an uploaded database file)
Write an email to send to investors
Problem solved.
No. The GPT kept looking into the PDF I uploaded with contact information on venture funds, forgetting to ask for me to upload a deck and rambling in its suggestions and funds to contact. All over the place.
Then I tinkered with the instructions, which is the way you control how the GPT behaves. I carefully told it in steps to do the above and felt confident. I also added some other very specific instructions to keep the answers focused. I mean, 3 steps.
How hard can it be?
As hard as getting a 5 year old to put on winter clothing apparently = Impossible.
Me explaining to my 5 year old son that he can’t wear shorts in the snow. Again.
Fast forward: Beta launch!
I don’t want to bore you with the details of getting that little **** to put on winter clothing.. I mean get that GPT to analyse a deck really well.
It took a while. It took blood, sweat, tears and I’ll be honest, it’s near impossible to figure out why the instructions give 100% different outputs when using the same setup, the same file and 0 changes. What the actual fuck. Toast in, shit out.
Here we are → Turbo.vc Beta: Try it!
But remember:
It only analyses text. Not images.
It’s rather inconsistent in the feedback, so uploading the same PDF twice might give very different results.
It’s indicative with the answers at best, if you are lucky that it understands your pitch…
It won’t perform better than talking and getting feedback from someone with solid startup or VC experience.
But it’s fun to try out and it actually did mention some things I had not thought of with our deck.
I wanna continue improving it in the coming weeks and I hope I can get it to be more consistent, but I’m honestly not sure I can make it super stabile right now.
Is GPT the future and did I just raise $$$ by using Turbo.vc?
It’s pretty incredible what you can create in a few hours. I mean, this idiot just did it.
It’s also crazy how all over the place the results are. I can’t explain in words how frustrating it is to get 100% different results with e-x-a-c-t-l-y the same inputs.
Some of the other GPTs I’ve tested in OpenAI’s GPT store have been very good I have to admit, so if you are smart unlike me, you can probably make it work way way better for your own custom GPT.
Right now it’s difficult for me to get it to be consistent even with very very strict guidelines, but improvements in AI are happening so incredibly fast and I wouldn’t be surprised if my dream of a “$$$-raising-GPT” comes true in 6 months.
For now, I’ll raise my money the old school way: using my massive brains!
Thanks for reading. I’ve missed this.