Artificial intelligence: what it is and why it concerns experts

This year, only 4 companies of IA generative have collected more than 370 million dollars (alrededor de 373 million euros at actual exchange rate) in capital, and 3 of them have reached the status of unicorn (company with value of more than 1,000 million) to break into the technological scene and attract the fascination of investors and consumers alike .

Las empresas —Cresta, Adept AI, Stability AI and Jasper— are promising in the field of generative AI, an emerging space whose valuations of 1,000 million dollars and massive rondas have contrasted with the general pessimism of the capital risk of 2022.

pero,What is generative AI? ¿Qué es lo que tiene a los inversores clamando por un lugar en las rondas al rojo vivo mientras luchan con las preoccupaciones éticas y de seguridad existentiales de la industria?

The popularity of generative AI

En pocas palabras, la IA generative se refers to the artificial intelligence that puede crear contenido.

“The model does not try to classify an image, por ejemplo, si es un cato un perro, o decir una prediction del tipo, oye, ¿cómo es tu retorno de clientes?”, cuenta Kanu Gulati, partner of Khosla Ventures . “Actually, se trata de generar contenido desde cero“.

The technological advances of the last years—in concrete terms of models of transformation and diffusion—have made generative AI possible. The models of IA refer to the training and display of algorithms in a set of data.

Los modelos transformeres entienden el contexto, lo que los hace ideales para generara un lenguaje humano lleno de matices. The diffusion models function by adding noise to the images before inverting the process to recreate the original image. Next, estos models pueden volver aplicar este proceso aprendido al ruido puro, eliminando lentamente el ruido hasta crear la imagen deseada. This process is the one that helps companies like Stability AI to generate images.

“No es una democratización, es una banalización del arte”: los artistas charge against the images generated by IA, aunque algunos les ven potential

An image generated by artificial intelligence.

The enthusiasm for generative AI began with users immersed in online communities dedicated to publishing images generated by AI, from the Shrek institute yearbook photo to a medieval tapestry representing the abduction of a cow by an extraterrestrial.

In recent times, startups have applied this technology to more practical cases.

In the field of images generated by AI, there are a lot of companies like OpenAI, Stability AI, Midjourney and Craiyon, while startups that generate text, like Jasper and, help writers create articles and blog headlines. Other startups produce video avatars, edit photos and videos, write codes and create customer responses.

Para Saam Motamedi, partner of the capital risk company Greylock, some of the most interesting advances come from multimodal AI, and AI that functions in multiple modes such as text, images and video, which, according to him, will “los trabajadores sean más productivos”.

Motamedi puso dinero detrás de esa apuesta invirtiendo en Adept AI, una startup que crea un “a system that can do any thing that a human can do in front of a computer”. In practice, this means that users can instruct the AI ​​to perform tasks that range from finding a house with 4 rooms to creating a column of benefits in Excel.

Ser más realist

Despite their optimism in this area, Gulati and Motamedi coincide in that the generative AI also poses a series of ethical and security problems.

Motamedi organizes its concerns in 3 groups: those who have a malicious actor, those who do not have a malicious actor and those who refer to the property.

In the first group, individuals can “poison” insidiously the training data to distort the result of the AI ​​or even use it to supplant others

“Hay una empresa en la que invertimos, Resemble AI, que realemente genera voces”, cuenta Bryan Rosenblatt, partner de Craft Ventures. “Dios no lo quiera, ver un mensaje del presidente diciento algo que han dicho: eso da miedo”.

Los actores de doblaje creen que la inteligenia artificial no acabará con su trabajo, pero asumen que traerá más precariedad al sector

Locución y doblaje con inteligenia artificial.

Even when there are no malicious actors, the biases in the data with which the AI ​​is trained can lead to biased results, for example, a model that creates a photo of a white man when asked to produce an image of a CEO , Cuenta Gulati.

In addition, judgments of great repercussion such as Thaler against Vidal raise doubts about who is the author of the art generated by AI.

Y aunque los inversores y los founderes assuran a los consumptives y los creadores que la IA se utilizará para aumentar —it’s time to replace it— las capacitas humanas, muchos siguen preguntándose si están en juego los puestos de trabajo de artists, vendadores y desarrolladores.

For Gulati, Motamedi and Rosenblatt, who have invested in this sector, the key is to find companies that incorporate security barriers in each step of their development process, such as the OpenAI content filter, and that actively create defense tools, como el detector de voz falsa de Resemble AI.

Para otros inversores, las preocupaciones se centran en la viability de las startups de las IA generativa como negocios centrados en la empresa financiadas con capital riesgo.

“Where I see the long-term budget in the company is normally around the financial director, the CISO, the human resources manager, the technology director, the sales vice president, the finance vice president,” says Anna Khan, a partner general de CRV. “No veo que la IA generativa hoy en día a través de sus primeros casos de uso exija ninguno de esos presupesuedos todavía”.

Khan adds that he is still not convinced that the products of generative AI are sufficiently convincing to win large-scale contracts and expansion of use, but he is attentive to the traction of products among skeptical non-technological companies.

Mientras tanto, los inversores esparan que tanto las empresas como los consumers sigan experimentando con esta tecnología.

Todo el mundo puede ser usuario, todo el mundo puede experimentar su magia y todo el mundo tiene ese momento de “aha”, says Rosenblatt. “La viralidad está incorporada”.

Leave a Comment

Your email address will not be published. Required fields are marked *