Luis García de la Fuente

Blog

Las alucinaciones son una feature, no un bug de los LLM

Las alucinaciones son una feature, no un bug. Este concepto se puede usar para crear nuevos negocios y productos. Si sólo tienes unas ideas vagas de lo que quieres hacer la IA te puede ayudar a cristalizarlas en propuestas de valor concretas.

Artificial Intelligence to Generate Business Ideas?

A blind LLM can create and refine images?

Researchers at MIT CSAIL found that large language models (LLMs) trained only on text data have an impressive understanding of visual concepts. By prompting LLMs to generate code for rendering images, the researchers collected a dataset of simple digital illustrations. Remarkably, the LLMs could iteratively improve these illustrations when prompted, demonstrating their robust visual knowledge gained from textual descriptions.…

HAL149 Mission Status: cambios en branding y aplicación online

La idea original de priorizar las interacciones directas con clientes no es casualidad. Teóricamente en un nuevo desarrollo tecnológico lo importante es vender primero a ‘los ricos’, porque sólo ellos pueden pagar el tiempo y recursos necesarios para financiar (al menos en parte) una nueva plataforma. Una vez que tengamos toda la información que necesitamos de lo que funciona, de lo que no, de lo que el cliente valora y cómo entregárselo aparecerá el producto: y nadie más podrá hacer nada parecido porque nadie ha tenido la voluntad o la capacidad de atravesar esta primera etapa.…

El mundo objetivo sin personas que lo imaginan, crean y recrean ¿qué es?. La conciencia es la única y primordial causa. Y todo el empeño del sistema de creencias y mentiras en el que vivimos tiene por objetivo hacernos creer que hay un mundo sólido y gigante fuera de nosotros, amenazante y caótico. Y que para poder navegarlo debemos ser protegidos (es decir, esclavizados) por burócratas, y todo tipo de parásitos.…

Guess what: AGI hysteria is a psyop operation

The AGI hysteria, predicting that the Terminator is coming for us all, is driven by psycho-sociopaths in the media. These same people would have burned anyone who dared to build an aeroplane centuries ago (“How dare you, God will punish us all for trying to fly like him!”).

The same level of absurdity is rampant today: “We have to be careful or some big language model will become sentient and its first decision will be to kill everyone”.…

The Digital Copyrights Nightmare

A very similar debate happened with copyrighted digital content, then with adblockers, and now with AI extracting ideas from content.

– People wanted to read free online copies of content with ads,
– Then people wanted to remove the ads to get the content,
– Now people just want to extract the ideas with AI from the content.
– Maybe in a few years people will want to get the ideas directly into their brains?…

Emergence at it finest: Use machines to understand how animals communicate

New algorithm discovers language just by watching videos. DenseAV, developed at MIT, learns to parse and understand the meaning of language just by watching videos of people talking, with potential applications in multimedia search, language learning, and robotics.

https://news.mit.edu/2024/denseav-algorithm-discovers-language-just-watching-videos-0611

Technology’s moving target

Everyone is confused and stressed about what to build on AI because everyone is focused on a moving target: technology.

Nobody is talking about what people really need. The real challenge is not about the AI models, the apps, the blockchain, etc., but about how to create value.

Technology is and will remain a moving target. It is only part of the equation: the fundamental part is people.…

¿Es la Economía un sistema disipativo?

El problema económico es básicamente un problema físico. Las economías humanas pueden ser estructuras disipativas no muy diferentes de las biológicas: plantas, animales, ecosistemas. Necesitan un aporte constante de energía.

A medida que la economía crece necesita cantidades crecientes de energía para funcionar. Por eso los negocios y los estados necesitan crecer para gestionar estas megaestructuras.…

Does AI have agency?

Agency in the context of cognitive science, particularly when considering the free energy principle, extends beyond just human decision-making and autonomy. It encompasses a broader understanding...

There are two different types of people in the world, those who want to know, and those who want to believe.

~ Nietzsche

La Simulación

El mundo es una proyección arquetípica de una realidad con más dimensiones.

Como la pantalla de tu ordenador, en la que se muestra una imágen interactiva de internet por la que puedes navegar e interaccionar en dos dimensiones.

Una imagen que es es fundamentalmente es luz e información. Información que se crea y re-crea instantaneamente en todo el mundo a través de millones de pantallas.…

Feasibility of Registering and Developing 1,000 Domains in Untapped Market

Q. I have compiled a list of almost 1,000 unregistered domains in my country’s TLD that contain valuable keywords (e.g. mortgage.tld, banks.tld, carinsurance.tld). I can register these for $5 each. With a team of 5 people, we can populate these sites with content over 2 months by working on 4 sites per day. The goal is not to monetize through ads due to extremely low CPMs, but rather to build authoritative sites and then sell them to businesses in those industries for around $1,000 each or get $20-50/month advertising deals.…

Humanoid robots: the new distraction for the plebs

Humanoid robots are another distraction for the plebs.

Now that AGI is fading from the news (who would have thought), the new kid in town to inflate the semiconductor bubble is… humanoid robots.

Stop following the BS and propaganda. Learn to focus on how AI can deliver real value to people and businesses. The future is about augmenting humans with AI, not Terminator wannabes.…

Poner un nombre nuevo en la cabeza de la gente es el primer paso para crear algo nuevo; como un contenedor al que hay dotar de significado. A continuación se trata de afinar el significado que se le da, más allá del éxito del negocio o valor tangible. Hacer a la gente participar en una revolución desde el principio, al tiempo que se genera empleo y valor.

Enhancing LLM Data Analysis through Specialized Tool Integration

Large language models like GPT-4 have shown incredible capabilities, but they fall short when it comes to rigorous data analysis, especially on large datasets. This shortcoming arises from their relatively small context window sizes that cannot accommodate full datasets, as well as the probabilistic nature of their text generation process which is ill-suited for precise statistical calculations.…

The 5-Step Feynman Technique with ChatGPT

The Feynman technique is recommended for rapid learning. We have a variation of this technique that involves five steps. Five Steps for Learning:

  • Select a Skill/Topic: Decide on the subject or skill you want to master.
  • Find Key Concepts: Use Chat GPT to get an overview of the topic and then break it down into key concepts.
  • Explain Like You’re Talking to a Child: The Feynman technique emphasizes explaining concepts as if to a child.

Privacy for the masses will soon be history

In the coming years, the Internet will become more centralised than ever before and will eventually become just another media system, like cable TV.

You will have issued your “internet access device” and it will only run on approved services and software that is fully monitored.

This is closer than it looks with mobile applications. And it will happen on the internet, using some kind of “event” as an excuse for ISPs/networks to block traffic.…

EU Passes the Artificial Intelligence Act

Attention, mortals and serfs! Behold, the Supreme Soviet has spoken and in its mighty wisdom has granted us yet another set of rules and expenses for our safety, the profit of our masters and the glory of democracy, equality and free markets.


The Artificial Intelligence Act (AI Act) is a regulation by the European Union to create a common legal framework for AI within the EU.…

AI Deception: How Tech Companies Are Fooling Us

En tecnología hay varios mundos o niveles de abstracción. En el mundo que se mueve la prensa y los “influencers” apenas hay valor para los negocios: la mayoría de la información se centra en los avances más recientes. Y éstos no siempre se pueden traducir en valor para las empresas.

Por otra parte para crear un producto no hace falta estar a la última: ésta es una de las grandes falacias que se alimenta desde los medios a todas horas.…

OpenAI transcribed over a million hours of YouTube videos to train GPT-4

All of these models are based on privatizing the commons, literally the whole of the internet.

This is a very different approach to Google, which basically lives on crediting authors and sending them traffic.

Also, if you ask a model to help you scrape a website, it’ll go on a ethics tirade about how questionable scraping is.…

AI Equivalence in Copilot

the request: “Please draw a room with absolutely no bananas in it”

is equivalent to …

the request: “Please draw a Zen room with some bananas in it”