Alejandro Piad Morffis (@alepiad) header
Alejandro Piad Morffis (@alepiad) profile picture

Democratizing knowledge one keystroke at a time. PhD in NLP, full-time professor, CS Department @UdeLaHabana. Co-founder @syalia_srl.

Havana, Cuba 🇨🇺
Joined September 2011
Photo by Alejandro Piad Morffis (@alepiad)Photo by Alejandro Piad Morffis (@alepiad)Photo by Alejandro Piad Morffis (@alepiad)Photo by Alejandro Piad Morffis (@alepiad)Photo by Alejandro Piad Morffis (@alepiad)Photo by Alejandro Piad Morffis (@alepiad)Photo by Alejandro Piad Morffis (@alepiad)Photo by Alejandro Piad Morffis (@alepiad)Photo by Alejandro Piad Morffis (@alepiad)Photo by Alejandro Piad Morffis (@alepiad)
Why are (artificial) neural networks so powerful? No, it's not because they are "biologically-inspired." That has little to do with it. It boils down to: - Their theoretical properties. - Their practical capacity to scale. Let me explain 👇
1
8
40
2,795
1) Neural networks are universal approximators This means that, theoretically, there is a big enough, super simple NN that can approximate any (sensible) mathematical function. (Now, that network is not the one you often want).
1
4
134
If you're the kind of nerd (like me) who prefers to chat with an LLM in the terminal instead of a website or app, check this project I'm working on. It's still pretty rough around the edges, but it knows Bash and it can mess with your system in fun and unexpected ways.
1
3
10
1,310
Here's a fun little LLM-based project I'm working on: github.com/apiad/lovelaice It's a bot that lives in your terminal and knows a bit of Bash, Python, and Linux. It can run commands and create files and maybe wipe out your hard drive if you're not looking. Take it for a spin?
2
8
858
LLMs cannot reason. Despite their impressive capabilities, all LLMs, including OpenAI o1, are still fundamentally limited by design constraints that make them incapable of true, open-ended reasoning. Let's break it down. 🧵 (1/5)
75
144
41
1,153
482,406
Reason 1: Stochastic Sampling. LLMs rely on probabilities to pick the next token. Even when you fix the temperature, randomness is still built into the language modeling paradigm. But logic is anything but random.
22
4
4
122
35,166
A few somewhat boring comments and predictions about generative AI in the short term.
1
2
12
1,949
Just as there was no reason to believe GPT-4 was an absolute hard limit, GPT-4o still isn't, but there is definitely diminishing returns in scale. At some point, rather sooner than later, bigger models *alone* won't give us that much of a qualitative improvement.
1
4
185
I'm writing a super practical ebook on using LLMs to build all sorts of cool stuff. Aimed at anyone with a minimum of coding experience, no previous knowledge of AI or NLP required. If you're interested, DM me and I'll send you a link to a 66% early access discount.
1
3
21
2,391
My best friend from college @dvdarias is the technical mastermind behind the first Cuban indie game. I've played some demos, and it's a masterpiece of storytelling, visuals, and gameplay. Super excited to get my hands on the final version. I bet you can't finish it before I do.
Three fates collide. One truth awaits. Saviorless finally breaches the veil on April 2nd, 2024 for Switch, PS5 & Steam. Will you be the Savior, the Hunter, or something else entirely? Wishlist now: saviorless.com #Saviorless #IndieGameDev
1
5
33
1,904
Hey folks! I'm writing the book on graph theory and algorithms you didn't know you needed. The introductory chapter is already online. Check it out! graphs.apiad.net
4
21
1
79
7,736
What do the World Wide Web, your brain, the corpus of scientific knowledge accumulated by all of humanity, the entire list of people you’ve ever met, and the city you live in have in common?
4
2
16
3,011
These are all very different types of things, from physical, to virtual, to social in nature, but they share a very important trait. They are all networks that establish relationships between some entities.
1
1
96
Yep, open-source is the only way to make AI safe and reliable. blog.apiad.net/p/mostly-harm…
The heretofore silent majority of AI scientists and engineers who - do not believe in AI extinction scenarios or - believe we have agency in making AI powerful, reliable, and safe and - think the best way to do so is through open source AI platforms NEED TO SPEAK UP !
1
2
1,694
I've been thinking and experimenting for a while on how to use generative AI to improve my writing productivity, without losing my voice and style. This is what I've converged to, a decently simple but effective workflow. open.techwriters.info/techwr…
1
1
12
1,811
New domain, same content. Here's a fresh startup article to update you on everything I've written in 2023. blog.apiad.net/start-here
2
1
9
1,515
The second issue of my subscriber-only newsletter on the state of AI is scheduled for next Monday. Why I think the future of AI is open-source. If you subscribe before, I'll give you a free 1-month upgrade so you can read all premium content. open.techwriters.info/apiad/…
4
1,126
Reductio ad ignoratum, the claim that because we don't know the exact answer to a given question, that answer might just be anything at all. Don't know what causes those lights? They could be aliens! Don't know what is consciousness? LLMs could be conscious!
5
2
1
19
2,277
If you flip a coin 20 times, it's extremely unlikely you'll get 20 heads. If 1 million people flip a coin 20 times, you're pretty certain someone will get 20 heads. If you only heard about those 1 in a million, you'd think they where vastly more talented than you.
5
54
5,295
I've been toying around with @monsterapis for a couple days and made this Telegram bot that will take your audio notes (in almost any language) and transcribe them for you. Take it for it a spin, and let me know what you think. It's free (for now). t.me/lovelaice_bot
1
7
889
A sublime insight of studying discrete math that I try to get my students to arrive at by themselves is the realization that theorems and algorithms are equivalent.
2
20
1,469
*At least when you have a constructive proof, which is often the case in discrete math.
6
460
First baby: - Honey!!! The baby is crying!!! - Oh my god, what's going on?! - Is she going to DIE!!!??? Second baby: - Honey, is that the baby crying? - Yeah... I think so. - Ah ok, then she's alive.
5
40
2,293
Katherine just turned one month old. Damn, time flies.
5
1
57
2,751
I've been playing with @monsterapis lately. Here's a sneak peek. So far, you can upload an audio file and get a transcription, plus a rewrite using Llama2 to improve grammar and spelling. You will need a MonsterAPI account. lovelaice.streamlit.app/
1
5
966
Next steps: I want to split the transcription into meaningful paragraphs, rewrite each paragraph separately, provide follow-up questions, and add a summary and section headers.
2
371