Connect with us

Artificial Intelligence

¿Máquinas con sentido común?

Published

on

máquinas con sentido común supnews
Imagen de Seanbatty en Pixabay

La creciente oleada de inteligencia artificial de los últimos tiempos está teniendo entre uno de sus retos, la integración del ‘sentido común’ a las máquinas a través del deep learning y otras tecnologías. El  aprendizaje profundo consiste en crear e implementar en los aparatos una red neuronal artificial que emula las funciones del cerebro humano. 

 

No obstante, muchos investigadores del ramo no están de acuerdo y alegan que existen mejores formas de aprendizaje robótico. Como la automatización de tareas en base a pronósticos y programas de lógica

 

Actualmente, el avance en este tema es vertiginoso. Los métodos de aprendizaje para las máquinas son cada vez más sofisticados. Aún así, el sentido común o la capacidad de discernir, tomando en cuenta elementos abstractos y en base a ello actuar e interactuar con el entorno, no es una capacidad fácil de integrar en un mecanismo artificial. Lo que se ha llevado a cabo hasta ahora, no es algo 100% espontáneo. Lo que sí se ha podido lograr es la simulación de reacciones y respuestas humanas por parte de robots.

 

Investigaciones y avances

 

Grandes empresas mundiales como Facebook, Apple, Microsoft y Google, han invertido importantes cantidades de dinero en el desarrollo de tecnologías de identificación facial, ubicación de patrones y preferencias de los usuarios y reconocimiento de voces y expresiones. 


TAMBIÉN TE PUEDE INTERESAR:

Inteligencia Artificial: El dilema entre seguridad y privacidad


Sin embargo, han sido numerosas las fallas técnicas, pues dentro de la inteligencia artificial el más mínimo error de programación puede traer resultados confusos e inesperados. Una máquina no entiende los conceptos de intromisión en la privacidad o de justicia y libertad.

 

Actualmente hay muchas iniciativas que tienen como misión principal la creación de ‘conciencia’ en los aparatos. En Seattle, Washington, el Instituto Allen trabaja en el Proyecto Alejandría con una inversión de más de 120 millones de dólares, para el desarrollo de sentido común en las máquinas. 

 

Por otra parte, la compañía Vicarious, con sede en San Francisco, California, con el apoyo de magnates tecnológicos como Elon Musk y Mark Zuckerberg, está desarrollando robots que imiten la capacidad humana de dominar múltiples funciones y cambiar de una actividad a otra como lo hacemos las personas. 

entido común supnews

Imagen de Jonny Lindner en Pixabay

El mismo Pentágono está colaborando con investigaciones en diversas universidades en el desarrollo de tecnologías que imiten artificialmente el razonamiento humano. Dentro de La Agencia de Proyectos de Investigación Avanzados de Defensa (DARPA) se creó el programa Machine Common Sense (MCS) que también está desarrollando una investigación para insertar el sentido común en máquinas inteligentes. En ella se incluyen conocimientos de el procesamiento del lenguaje natural, la comprensión cognitiva y el aprendizaje profundo.

 

El futuro y las máquinas con sentido común

 

Desde hace décadas la ciencia ficción nos planteaba muchas opciones en avances tecnológicos que ya forman parte de nuestra cotidianidad. Un claro ejemplo de ello son las video llamadas. En cuanto a la creación de mecanismos artificiales con características y reacciones humanas, el camino ya está siendo recorrido. 

 

La conocida robot humanoide Sofía ha sorprendido al planeta entero, con su capacidad de contestar de forma casi espontánea en múltiples exposiciones y entrevistas. No obstante, sigue siendo un prototipo con mucho por mejorar y con carencia de numerosas respuestas. 

Las proyecciones a futuro de esta realidad están divididas entre adeptos y detractores. Las fantasías catastróficas basadas en la cinematografía, donde los robots con conciencia propia toman el mando de la humanidad en detrimento de las personas. Así como las utópicas que presentan una vida cómoda y placentera en la que los robots se encargan de ejecutar las labores tediosas y pesadas del día a día son extremas. 

 

Lo que sí es cierto es que los avances en la creación de máquinas con sentido común son impresionantes y en muchos casos, también espeluznantes. Y que esta realidad podría encontrarse a la vuelta de la esquina.

 

¿Cuál es tu percepción en relación a la inteligencia artificial y a los robots con conciencia propia? Compártenos tu opinión.

Artificial Intelligence

Facebook blames COVID-19 for reduced action on suicide, self-injury, and child exploitation content

Published

on

facebook-blames-covid-19-for-reduced-action-on-suicide,-self-injury,-and-child-exploitation-content

Facebook says that COVID-19 has hindered its ability to remove posts about suicide, self-injury, and child nudity and sexual exploitation.

The social media giant said the decision to send content reviewers home in March had forced it to rely more heavily on tech to remove violating content.

As a result, the firm says it took action on 911,000 pieces of content related to suicide and self-injury in the second quarter of this year — just over half the number of the previous quarter.

On Instagram, the number dropped even further, from 1.3 million pieces of content in Q1 to 275,000 in Q2. Meanwhile, action on Instagram content that sexually exploits or endangers children decreased from 1 million to 479,400.

“With fewer content reviewers, we took action on fewer pieces of content on both Facebook and Instagram for suicide and self-injury, and child nudity and sexual exploitation on Instagram,” said Guy Rosen, Facebook‘s VP of Integrity, in a blog post today.

[Read: Social media firms will use more AI to combat coronavirus misinformation, even if it makes more mistakes]

Facebook said that stretched human resources had also reduced the number of appeals it could offer. In addition, the firm claimed that its focus on removing of harmful content meant it couldn’t calculate the prevalence of violent and graphic content in its latest community standards report.

More human moderation needed

Facebook did report some improvements in its AI moderation efforts. The company said the proactive detection rate for hate speech on Facebook had increased from 89% to 95%. This led it to take action on 22.5 million pieces of violating content, up from the 9.6 million in the previous quarter.

Instagram‘s hate speech detection rate climbed even further, from 45% to 84%, while actioned content rose from 808,900 to 3.3 million.

Rosen said the results show the importance of  human moderators:

Today’s report shows the impact of COVID-19 on our content moderation and demonstrates that, while our technology for identifying and removing violating content is improving, there will continue to be areas where we rely on people to both review content and train our technology.

In other Facebook news, the company today announced new measures to stop publishers backed by political organizations from running ads disguised as news. Under the new policy, news Pages with these affiliations will be banned from Facebook News. They’ll also lose access to news messaging on the Messenger Business Platform or the WhatsApp business API.

With the US election season approaching, it’s gonna be a busy few months for Facebook‘s content moderation team.

Published August 11, 2020 — 18:21 UTC

Thomas Macaulay

Thomas Macaulay

August 11, 2020 — 18:21 UTC

Continue Reading

Artificial Intelligence

Pinterest improves and expands its skin tone search feature

Published

on

pinterest-improves-and-expands-its-skin-tone-search-feature

Pinterest is upgrading its skin tone search feature, which uses machine vision to sort pins in the site’s beauty category by skin tone. The feature launched in the US in 2018 and is now available in the UK, Canada, Ireland, Australia, and New Zealand as well.

The feature is designed to make it easier for users to find content relevant to them, says Pinterest. It’s a common problem in the search world that certain queries default to show white faces. By giving users the option to refine their searches based on skin tones, Pinterest says it helps users find they content they want to see.

The feature is now more prominent when users are searching for content and delivers more accurate results, says Pinterest. The company offers searches like “grey hair on dark skin women,” “blonde hair color ideas for fair skin blue eyes,” and “soft natural makeup for Black women“ as examples of the sort of fine-grained results the feature can deliver.

Pinterest’s Try On feature lets users try on lipstick shades in AR.
Image: Pinterest

Search by skin tone is also now integrated into the company’s augmented reality Try On feature, which lets users search for lipstick shades and try them on in AR. This feature is currently only available in the US but is launching in the UK “in the coming months.”

Continue Reading

Artificial Intelligence

UK court rules police use of facial recognition was ‘unlawful’

Published

on

uk-court-rules-police-use-of-facial-recognition-was-‘unlawful’

British police used facial recognition unlawfully, the Court of Appeal ruled today, in a landmark decision that could have a big impact on the technology’s use in the UK.

The judgment stems from a complaint by Cardiff resident Ed Bridges, who said police had scanned his face while he was Christmas shopping, and again when he was at a protest.

Bridges argued that South Wales Police (SWP) had breached his right to privacy, as well as equality and data protection laws. But last September, the UK‘s Supreme Court ruled against him, claiming cops had followed the relevant rules and met the requirements of the Human Rights Act.

Bridges appealed the decision, arguing that SWP’s actions were akin to taking fingerprints or DNA without consent. Bridges was supported by human rights group Liberty, which says the case is the world’s first legal challenge to police use of automated facial recognition (AFR).

[Read: Clearview AI can be fun — if you’re dirty, stinking rich]

Today, the Court of Appeal agreed that police had violated his right to privacy, as well as data protection and equality laws.

The judges said that “too much discretion is currently left to individual police officers,” and that SWP had “never sought to satisfy themselves, either directly or by independent verification, that the software program does not have an unacceptable bias on grounds of race or sex.”

Bridges said he was “delighted” with the decision:

This technology is an intrusive and discriminatory mass surveillance tool. For three years now South Wales Police has been using it against hundreds of thousands of us, without our consent and often without our knowledge. We should all be able to use our public spaces without being subjected to oppressive surveillance.

Future implications for facial recognition

The judges called for changes to the framework that regulates AFR. These could involve amendments to local policy documents, such as those operated by South Wales Police, or to the national Surveillance Camera Code of Practice.

However, they didn’t rule that primary legislation — the main laws passed in the UK — were required to regulate AFR in the same way as DNA or fingerprints.

“Instead, the Court has identified the relatively modest changes to the policy framework that are needed in order that live AFR can continue to be used,” said Anne Studd, a senior lawyer at 5 Essex Court who specializes in police law.

“It is noteworthy that this case arose in the course of a pilot of the system by South Wales Police – as part of that trial, through a co-operative and consensual process by which the issues were brought before the Court, the police service has been able to obtain a very helpful decision that maps the way ahead.”

South Wales Police and London’s Metropolitan Police were reportedly the only forces in the UK using AFR. Liberty is now calling for them to stop using the tech entirely.

Published August 11, 2020 — 11:10 UTC

Continue Reading

Trending

English
Spanish English