La burbuja de la IA

[muyuu]

https://xcancel.com/slow_developer/status/2009445018763673960

Terence Tao says the current business model of big AI companies relies on hype to justify their massive valuations

The situation is similar to the dot-com boom, where the tech was real, but the early promises were inflated

"hopefully we'll reach a more realistic equilibrium in a year or two"

[Benzino Napaloni]

https://xcancel.com/slow_developer/status/2009445018763673960

Terence Tao says the current business model of big AI companies relies on hype to justify their massive valuations

The situation is similar to the dot-com boom, where the tech was real, but the early promises were inflated

"hopefully we'll reach a more realistic equilibrium in a year or two"

Hace 25 años quedaron en pie Google, Amazon, y otras con valor real. Cierto. Pero la purga fue muy dolorosa.

Hay quien estima que la burbuja de la IA es 17 veces mayor que la de las puntocom. Y 4 veces mayor que la financiera de hace 20 años.

Llegar a un equilibrio ahora (que va a suceder sí o también) implica pegar un pedo que lo van a oír los marcianos.

[tomasjos]

Inteligencia artificial, negligencia real: la abogacía se la juega con el algoritmo https://share.google/SitxE4UmqpgYYGnMD

Al parecer los abogados usan IA para redactar sus alegatos y la jurisprudencia y está alucina

[muyuu]

un poco más relacionado con el otro hilo, pero lo cito aquí

C-Suite son ejecutivos nivel CEO/CFO/CTO

[muyuu]

https://www.gizmochina.com/2026/01/21/openai-wants-a-cut-of-your-profits-inside-its-new-royalty-based-plan-and-other-business-models/

OpenAI Wants a Cut of Your Profits: Inside Its New Royalty-Based Plan and Other Business Models - Gizmochina
Joane
3 - 4 minutes

OpenAI is preparing for a major shift in how it makes money. The company, best known for ChatGPT, is moving beyond relying only on subscription fees. While paid plans will continue, OpenAI is now exploring additional revenue models such as licensing, profit-sharing, advertising, and outcome-based payments. This change reflects the company’s need to fund massive computing costs while supporting growing global demand for AI services.
Royalty and Outcome-Based Pricing

OpenAI CFO Sarah Friar has outlined a future where the company earns money only when its customers succeed. In this model, OpenAI could receive royalties or licensing fees tied directly to business outcomes. For example, if a pharmaceutical company uses OpenAI tools to develop a successful drug, OpenAI could take a small percentage of the drug’s sales. This approach aligns incentives, as OpenAI benefits only when customers generate real value.
The “Rubik’s Cube” Business Strategy

Friar describes OpenAI’s strategy as a “Rubik’s Cube” of business models. Each side of the cube represents different combinations of technology, pricing, products, and markets. This metaphor highlights flexibility and rapid change. Unlike its early days, when OpenAI depended on one cloud provider, one chip partner, and one main product, the company now operates with multiple partners, products, and pricing structures.
Expanding Products and Partnerships

OpenAI has significantly expanded its portfolio. Alongside ChatGPT for consumers and businesses, it now offers tools like Sora for video generation, enterprise AI platforms, industry-specific solutions, and scientific research systems. The company also works with multiple cloud providers and chip manufacturers, reducing dependence on a single partner.
Computing Power Drives Revenue

Demand for OpenAI’s services is extremely high, but growth is limited by available computing power. Over the past two years, OpenAI’s revenue has grown nearly tenfold, closely tracking its expansion in computing capacity. To support this growth, the company has committed to massive infrastructure deals with partners like Oracle and AMD.
What Lies Ahead

OpenAI is also testing advertising for free users and exploring e-commerce features. The long-term goal is to make AI as reliable as basic infrastructure, similar to electricity. As AI agents move from experiments to core business tools, OpenAI believes diversified revenue models are essential to sustain its mission and meet future demand.

Qué cachondos son. Si salen por ahí les va a usar Rita, con lo bien que funcionan los modelos abiertos.

La jeta que se gastan intentando cobrar por los resultados cuando no han pagado un duro por los datos que han acopiado y usado en el entrenamiento de sus modelos.

[muyuu]

el modelo abierto chinorris que está dando que hablar esta semana (pendientes de Qwen que sacará otro pronto)

[pollo]

el modelo abierto chinorris que está dando que hablar esta semana (pendientes de Qwen que sacará otro pronto)

No hay modelo de negocio. Esto era trivial de ver.
Luego sacarán otro, y otro y otro, y no tendrá sentido pagar por ello, mucho menos cuando ni siquiera hace lo que se vende por el marketing.

[muyuu]

[...]

No hay modelo de negocio. Esto era trivial de ver.
Luego sacarán otro, y otro y otro, y no tendrá sentido pagar por ello, mucho menos cuando ni siquiera hace lo que se vende por el marketing.

el modelo de negocio depende casi por completo de que los modelos cerrados que requieren terabytes de RAM sean insustituíbles por modelos que sean utilizables a mediana escala

me hace gracia que los chinorris básicamente les están pirateando el entrenamiento a los grandes por "destilación" y por mucho que lloren, openAI, Google, Anthropic etc todos han acopiado información sin pedir permiso a nadie

he probado Kimi K2.5 en local (hace falta bastante equipo para hacerlo, pero con dos Mac Studio Ultra M3 que cuestan ~£9000 se puede) y si le dices "hi" te responde "hi, I'm Claude ..." y te suelta una perorate típica de Perplexity


imagino que pronto empezarán con un modelo de anuncios tipo Google sacándote toda la información que puedan, mientras te perjuran que respetan tu privacidad, no les veo otra salida

viendo los resultados que consigo en local con mis propias máquinas, no falta mucho para que casi nadie se beneficie de usar modelos cerrados aparte de modelos de integración, interfaz, que la gente se empeñe en usar un app y regale su información etc... cosas que no descarto

ahora mismo tengo una basada en Strix Halo que cuesta unos 2000 euros yo creo que con diferencia es la mejor en calidad-precio, otra basada en 4x nVidia 3090s, y los 2 mac Studio de 512GB conectados con Thunderbolt 5

con la de 2000 euros te basta para casi toda la usabilidad real que estoy viendo ahora mismo

un futurible es que casi nadie se moleste en comprarse ordenador incluso aunque los ordenadores de 128GB p'arriba de memoria unificada y con NPUs sean asequibles, sino que tengan móviles muy cerrados y android o lo que venga, y paguen subscripciones para que Google o Apple o suputamadre les lleven la vida

es el futurible más triste que se me ocurre pero lo veo muy plausible, y de ahí pueden salir unas cuantas compañías grandes que se lucren de vender la información de la gente como Google y Meta, y ahora también Microsoft y Apple

tristemente no me parece que sea para nada descabellado y encaja con los que son los modelos de negocios más rentables del mundo mundial actualmente, y los más depredadores también

nada de eso es incompatible con que quiebren todas o casi todas y que se quede Google o nVidia con todo lo gordo, por ejemplo

[muyuu]

https://www.windowscentral.com/microsoft/windows-11/microsoft-is-reevaluating-its-ai-efforts-on-windows-11-plans-to-reduce-copilot-integrations-and-evolve-recall

You won: Microsoft is walking back Windows 11’s AI overload — scaling down Copilot and rethinking Recall in…
In an exclusive report from Windows Central, Microsoft is mulling pulling back its Windows 11 AI push with a major Copilot and Recall rethink.

By Zac Bowden

4 min. read

It’s fair to say that Windows 11’s recent endeavour into AI hasn’t gone down well with its most passionate users. It started in 2024 with the unveiling of Windows Recall, which was met with such backlash that Microsoft was forced to postpone it by an entire year while it addressed major security and privacy flaws.

It seems like things have been downhill since. In the last year, Microsoft has taken every opportunity to enshittify Windows 11 by placing Copilot buttons wherever it can across in-box apps like File Explorer and Notepad, even if the implementation is poor or unnecessary.


This has soured Microsoft’s AI efforts in the eyes of many Windows users, resulting in major pushback online and adding to the overall negative sentiment around Windows 11. This came to a head in November, when Windows president Pavan Davuluri tweeted that Windows would evolve into an agentic OS, spawning thousands of overwhelmingly negative replies rejecting this plan.


It appears this moment of pushback has resonated with internal teams: According to people familiar with Microsoft’s plans, the company is now reevaluating its AI strategy on Windows 11 and plans changes to streamline or even remove certain AI features where they don’t make sense.

Details around how the company is going about this remain light, but sources say Copilot integrations like those found in Notepad and Paint are under review. This may result in Microsoft removing certain Copilot integrations from these apps, or at the very least removing the Copilot branding and pivoting to a more streamlined experience.

New Welcome Screen in Notepad detailing recent updates

Copilot features in Notepad? It's preposterous.
(Image credit: Windows Central)

I’m also told that Microsoft has paused work on any additional Copilot buttons for in-box apps, at least for now. While I don’t expect this pause to be permanent, it does sound like Microsoft plans to be more tactful and deliberate in where these Copilot buttons and integrations will appear going forward.

Windows Recall is another AI experience that I’m told is under review. Sources tell me that Microsoft believes that Recall, in its current implementation, has failed, though I understand the company is exploring ways to evolve the concept rather than scrap it entirely, possibly dropping the Recall name in the process, though this is unconfirmed.

Other AI initiatives, such as Semantic Search, Agentic Workspace, Windows ML, and Windows AI APIs, are continuing ahead as planned. Microsoft believes that these under-the-hood AI efforts are still important for app developers and users, positioning Windows as a viable contender amongst other OS’s that are also building AI frameworks into their platforms.

The company is shifting away from ‘AI everywhere’ and toward features that actually make sense for Windows users.

The good news is that it's clear Microsoft has heard the feedback around its heavy-handedness when it comes to Copilot buttons in Windows apps. The company is stepping back to readjust how best to implement these AI integrations across the OS, hopefully resulting in a more meaningful and useful AI experience on the platform, rather than haphazardly adding the Copilot icon to every UI surface it can.

This effort is likely part of Microsoft's overall effort to "fix" Windows 11 this year. I understand that the company is moving quickly to begin shipping meaningful changes that are designed to signal to customers that it is listening to feedback, and streamlining where Copilot shows up across in-box apps would be a strong place to start.