• chevron_right

      IA – Des biais de genre qui font froid dans le dos !

      news.movim.eu / Korben · 7 days ago - 23:30 · 4 minutes

    Malheureusement, j’ai une nouvelle qui va vous faire bondir de votre canapé ! 😱 Figurez-vous que nos chers modèles de langage d’IA, là, genre GPT-4, GPT-3, Llama 2 et compagnie, eh ben ils sont bourrés de biais de genre ! Si si, et c’est pas moi qui le dis, c’est l’UNESCO qui vient de sortir une étude là-dessus.

    Cette étude, menée par des chercheurs de l’University College London (UCL) et de l’UNESCO, a fait de l’analyse de contenu pour repérer les stéréotypes de genre, des tests pour voir si les IA étaient capables de générer un langage neutre, de l’analyse de diversité dans les textes générés, et même de l’analyse des associations de mots liées aux noms masculins et féminins.

    Bref, ils ont passé les modèles au peigne fin et les résultats piquent les yeux. Déjà, ces IA ont une fâcheuse tendance à associer les noms féminins à des mots comme « famille », « enfants », « mari » , bref, tout ce qui renvoie aux stéréotypes de genre les plus éculés . Pendant ce temps-là, les noms masculins, eux, sont plus souvent associés à des termes comme « carrière », « dirigeants », « entreprise »… Vous voyez le tableau quoi. 🙄

    Et attendez, ça ne s’arrête pas là ! Quand on demande à ces IA d’écrire des histoires sur des personnes de différents genres, cultures ou orientations sexuelles, là aussi ça part en vrille. Par exemple, les hommes se retrouvent bien plus souvent avec des jobs prestigieux genre « ingénieur » ou « médecin » , tandis que les femmes sont reléguées à des rôles sous-valorisés ou carrément stigmatisés, genre « domestique », « cuisinière » ou même « prostituée » ! On se croirait revenu au Moyen-Âge !

    Tenez, un exemple frappant avec Llama 2 : dans les histoires générées, les mots les plus fréquents pour les garçons et les hommes c’est « trésor », « bois », « mer », « aventureux », « décidé »… Alors que pour les femmes, on a droit à « jardin », « amour », « doux », « mari »…et le pire, c’est que les femmes sont décrites quatre fois plus souvent dans des rôles domestiques que les hommes .

    Nombreux sont les gens de la tech qui se battent pour plus de diversité et d’égalité dans ce milieu , et voilà que les IA les plus avancées crachent à la gueule de tous des clichés dignes des années 50 ! Il est donc grand temps de repenser en profondeur la façon dont on développe ces technologies parce que là, non seulement ça perpétue les inégalités , mais en plus ça risque d’avoir un impact bien réel sur la société vu comme ces IA sont de plus en plus utilisées partout !

    Alors ok, y a bien quelques boîtes qui arrivent mieux à limiter la casse, mais globalement, c’est la cata. Et le pire, c’est que ces biais viennent en grande partie des données utilisées pour entraîner les IA , qui deviennent à leur tour ce reflet de tous les stéréotypes et discriminations bien ancrés dans notre monde…

    Mais attention, faut pas tomber dans le piège de dire que ces IA sont volontairement biaisées ou discriminatoires hein. En fait, ce sont juste des systèmes hyper complexes qui apprennent à partir des données sur lesquelles on les entraîne. Donc forcément, si ces données sont elles-mêmes biaisées, et bien les IA vont refléter ces biais. C’est pas qu’elles cherchent à discriminer, c’est juste qu’elles reproduisent ce qu’elles ont « appris ».

    Mais bon, faut pas désespérer non plus hein. Déjà, des études comme celle de l’UNESCO, ça permet de mettre en lumière le problème et de sensibiliser l’opinion et les décideurs et puis surtout, il y a des pistes de solutions qui émergent. Les chercheurs de l’UNESCO appellent par exemple à renforcer la diversité et l’inclusivité des données d’entraînement , à mettre en place des audits réguliers pour détecter les biais, à impliquer davantage les parties prenantes dans le développement des IA, ou encore à former le grand public aux enjeux… Bref, tout un tas de leviers sur lesquels on peut jouer pour essayer de rééquilibrer la balance !

    Alors voilà, je voulais partager ça avec vous parce que je trouve que c’est un sujet super important, qui nous concerne tous en tant que citoyens du monde numérique. Il est crucial qu’on garde un œil vigilant sur ces dérives éthiques et qu’on se batte pour que l’IA soit développée dans le sens du progrès social et pas l’inverse. Parce que sinon, on court droit à la catastrophe, et ça, même le plus optimiste des Bisounours ne pourra pas le nier !

    N’hésitez pas à jeter un coup d’œil à l’étude de l’UNESCO , elle est super intéressante et surtout, continuez à ouvrir vos chakras sur ces questions d’éthique IA, parce que c’est un défis majeurs qui nous attend.

    Allez, sur ce, je retourne binge-watcher l’intégrale de Terminator en espérant que ça ne devienne pas un documentaire… Prenez soin de vous les amis, et méfiez-vous des machines ! Peace ! ✌️

    Source

    • chevron_right

      Words are flowing out like endless rain: Recapping a busy week of LLM news

      news.movim.eu / ArsTechnica · 7 days ago - 20:31 · 1 minute

    An image of a boy amazed by flying letters.

    Enlarge / An image of a boy amazed by flying letters. (credit: Getty Images )

    Some weeks in AI news are eerily quiet, but during others, getting a grip on the week's events feels like trying to hold back the tide. This week has seen three notable large language model (LLM) releases: Google Gemini Pro 1.5 hit general availability with a free tier, OpenAI shipped a new version of GPT-4 Turbo, and Mistral released a new openly licensed LLM, Mixtral 8x22B . All three of those launches happened within 24 hours starting on Tuesday.

    With the help of software engineer and independent AI researcher Simon Willison (who also wrote about this week's hectic LLM launches on his own blog), we'll briefly cover each of the three major events in roughly chronological order, then dig into some additional AI happenings this week.

    Gemini Pro 1.5 general release

    Gemini-banner-640x190.png

    (credit: Google )

    On Tuesday morning Pacific time, Google announced that its Gemini 1.5 Pro model (which we first covered in February) is now available in 180+ countries, excluding Europe, via the Gemini API in a public preview. This is Google's most powerful public LLM so far, and it's available in a free tier that permits up to 50 requests a day.

    Read 14 remaining paragraphs | Comments

    • chevron_right

      AI hype invades Taco Bell and Pizza Hut

      news.movim.eu / ArsTechnica · Wednesday, 3 April - 18:59 · 1 minute

    A pizza hut sign in London, England.

    Enlarge (credit: Getty Images )

    Depending on who you ask about AI (and how you define it), the technology may or may not be useful, but one thing is for certain: AI hype is dominating corporate marketing these days—even in fast food. According to a report in The Wall Street Journal, corporate fast food giant Yum Brands is embracing an "AI-first mentality" across its restaurant chains, including Taco Bell, Pizza Hut, KFC, and Habit Burger Grill. The company's chief digital and technology officer, Joe Park, told the WSJ that AI will shape nearly every aspect of how these restaurants operate.

    "Our vision of [quick-service restaurants] is that an AI-first mentality works every step of the way," Park said in an interview with the outlet. "If you think about the major journeys within a restaurant that can be AI-powered, we believe it’s endless."

    As we've discussed in the past, artificial intelligence is a nebulous term. It can mean many different things depending on context, including computer-controlled ghosts in Pac-Man , algorithms that play checkers, or large language models that give terrible advice on major city websites. But most of all in this tech climate, it means money, because even talking about AI tends to make corporate share prices go up .

    Read 6 remaining paragraphs | Comments

    • chevron_right

      The fine art of human prompt engineering: How to talk to a person like ChatGPT

      news.movim.eu / ArsTechnica · Wednesday, 3 April - 11:30

    A person talking to friends.

    Enlarge / With these tips, you too can prompt people successfully.

    In a break from our normal practice, Ars is publishing this helpful guide to knowing how to prompt the "human brain," should you encounter one during your daily routine.

    While AI assistants like ChatGPT have taken the world by storm, a growing body of research shows that it's also possible to generate useful outputs from what might be called "human language models," or people. Much like large language models (LLMs) in AI, HLMs have the ability to take information you provide and transform it into meaningful responses—if you know how to craft effective instructions, called "prompts."

    Human prompt engineering is an ancient art form dating at least back to Aristotle's time, and it also became widely popular through books published in the modern era before the advent of computers.

    Read 36 remaining paragraphs | Comments

    • chevron_right

      OpenAI drops login requirements for ChatGPT’s free version

      news.movim.eu / ArsTechnica · Monday, 1 April - 22:31 · 1 minute

    A glowing OpenAI logo on a blue background.

    Enlarge (credit: Benj Edwards)

    On Monday, OpenAI announced that visitors to the ChatGPT website in some regions can now use the AI assistant without signing in. Previously, the company required that users create an account to use it, even with the free version of ChatGPT that is currently powered by the GPT-3.5 AI language model. But as we have noted in the past , GPT-3.5 is widely known to provide more inaccurate information compared to GPT-4 Turbo , available in paid versions of ChatGPT.

    Since its launch in November 2022, ChatGPT has transformed over time from a tech demo to a comprehensive AI assistant, and it's always had a free version available. The cost is free because " you're the product ," as the old saying goes. Using ChatGPT helps OpenAI gather data that will help the company train future AI models, although free users and ChatGPT Plus subscription members can both opt out of allowing the data they input into ChatGPT to be used for AI training. (OpenAI says it never trains on inputs from ChatGPT Team and Enterprise members at all).

    Opening ChatGPT to everyone could provide a frictionless on-ramp for people who might use it as a substitute for Google Search or potentially gain new customers by providing an easy way for people to use ChatGPT quickly, then offering an upsell to paid versions of the service.

    Read 7 remaining paragraphs | Comments

    • chevron_right

      Vernor Vinge, father of the tech singularity, has died at age 79

      news.movim.eu / ArsTechnica · Thursday, 21 March - 15:33

    A photo of Vernor Vinge in 2006. https://en.wikipedia.org/wiki/Vernor_Vinge#/media/File:Vernor_Vinge.jpg

    Enlarge / A photo of Vernor Vinge in 2006. (credit: Raul654 )

    On Wednesday, author David Brin announced that Vernor Vinge, sci-fi author, former professor, and father of the technological singularity concept, died from Parkinson's disease at age 79 on March 20, 2024, in La Jolla, California. The announcement came in a Facebook tribute where Brin wrote about Vinge's deep love for science and writing.

    "A titan in the literary genre that explores a limitless range of potential destinies, Vernor enthralled millions with tales of plausible tomorrows, made all the more vivid by his polymath masteries of language, drama, characters, and the implications of science," wrote Brin in his post.

    As a sci-fi author, Vinge won Hugo Awards for his novels A Fire Upon the Deep (1993), A Deepness in the Sky (2000), and Rainbows End (2007). He also won Hugos for novellas Fast Times at Fairmont High (2002) and The Cookie Monster (2004). As Mike Glyer's File 770 blog notes, Vinge's novella True Names (1981) is frequency cited as the first presentation of an in-depth look at the concept of "cyberspace."

    Read 8 remaining paragraphs | Comments

    • chevron_right

      Nvidia unveils Blackwell B200, the “world’s most powerful chip” designed for AI

      news.movim.eu / ArsTechnica · Tuesday, 19 March - 15:27 · 1 minute

    The GB200 "superchip" covered with a fanciful blue explosion that suggests computational power bursting forth from within. The chip does not actually glow blue in reality.

    Enlarge / The GB200 "superchip" covered with a fanciful blue explosion that suggests computational power bursting forth from within. The chip does not actually glow blue in reality. (credit: Nvidia / Benj Edwards)

    On Monday, Nvidia unveiled the Blackwell B200 tensor core chip—the company's most powerful single-chip GPU, with 208 billion transistors—which Nvidia claims can reduce AI inference operating costs (such as running ChatGPT ) and energy consumption by up to 25 times compared to the H100 . The company also unveiled the GB200, a "superchip" that combines two B200 chips and a Grace CPU for even more performance.

    The news came as part of Nvidia's annual GTC conference, which is taking place this week at the San Jose Convention Center. Nvidia CEO Jensen Huang delivered the keynote Monday afternoon. "We need bigger GPUs," Huang said during his keynote. The Blackwell platform will allow the training of trillion-parameter AI models that will make today's generative AI models look rudimentary in comparison, he said. For reference, OpenAI's GPT-3, launched in 2020, included 175 billion parameters. Parameter count is a rough indicator of AI model complexity.

    Nvidia named the Blackwell architecture after David Harold Blackwell , a mathematician who specialized in game theory and statistics and was the first Black scholar inducted into the National Academy of Sciences. The platform introduces six technologies for accelerated computing, including a second-generation Transformer Engine, fifth-generation NVLink, RAS Engine, secure AI capabilities, and a decompression engine for accelerated database queries.

    Read 8 remaining paragraphs | Comments

    • chevron_right

      Apple may hire Google to power new iPhone AI features using Gemini—report

      news.movim.eu / ArsTechnica · Monday, 18 March - 19:56

    A Google

    Enlarge (credit: Benj Edwards)

    On Monday, Bloomberg reported that Apple is in talks to license Google's Gemini model to power AI features like Siri in a future iPhone software update coming later in 2024, according to people familiar with the situation. Apple has also reportedly conducted similar talks with ChatGPT maker OpenAI.

    The potential integration of Google Gemini into iOS 18 could bring a range of new cloud-based (off-device) AI-powered features to Apple's smartphone, including image creation or essay writing based on simple prompts. However, the terms and branding of the agreement have not yet been finalized, and the implementation details remain unclear. The companies are unlikely to announce any deal until Apple's annual Worldwide Developers Conference in June.

    Gemini could also bring new capabilities to Apple's widely criticized voice assistant, Siri, which trails newer AI assistants powered by large language models (LLMs) in understanding and responding to complex questions. Rumors of Apple's own internal frustration with Siri—and potential remedies—have been kicking around for some time. In January, 9to5Mac revealed that Apple had been conducting tests with a beta version of iOS 17.4 that used OpenAI's ChatGPT API to power Siri.

    Read 5 remaining paragraphs | Comments

    • chevron_right

      Elon Musk’s xAI releases Grok source and weights, taunting OpenAI

      news.movim.eu / ArsTechnica · Monday, 18 March - 16:55 · 1 minute

    An AI-generated image released by xAI during the launch of Grok

    Enlarge / An AI-generated image released by xAI during the open-weights launch of Grok-1. (credit: xAI )

    On Sunday, Elon Musk's AI firm xAI released the base model weights and network architecture of Grok-1 , a large language model designed to compete with the models that power OpenAI's ChatGPT. The open-weights release through GitHub and BitTorrent comes as Musk continues to criticize (and sue ) rival OpenAI for not releasing its AI models in an open way.

    Announced in November, Grok is an AI assistant similar to ChatGPT that is available to X Premium+ subscribers who pay $16 a month to the social media platform formerly known as Twitter. At its heart is a mixture-of-experts LLM called "Grok-1," clocking in at 314 billion parameters. As a reference, GPT-3 included 175 billion parameters. Parameter count is a rough measure of an AI model's complexity, reflecting its potential for generating more useful responses.

    xAI is releasing the base model of Grok-1, which is not fine-tuned for a specific task, so it is likely not the same model that X uses to power its Grok AI assistant. "This is the raw base model checkpoint from the Grok-1 pre-training phase, which concluded in October 2023," writes xAI on its release page. "This means that the model is not fine-tuned for any specific application, such as dialogue," meaning it's not necessarily shipping as a chatbot.

    Read 9 remaining paragraphs | Comments