The King’s Institute for AI discusses the growing GenAI use in education and the risks of false information disguised by eloquent writing. The post Are we prepared for the looming shadow of AI hallucination? appeared first on Innovation News Network. Source link
Hallucinations — the lies generative AI models tell, basically — are a big problem for businesses looking to integrate the technology into their operations. Because models have no real intelligence and are simply predicting words, images, speech, music and other data according to a private schema, they sometimes get it wrong. Very wrong. In a recent piece in The Wall Street Journal, a source recounts an instance where Microsoft’s generative AI invented meeting attendees and implied that conference calls were about subjects that weren’t actually discussed on the call. As I wrote a while ago, hallucinations may be an unsolvable problem with today’s transformer-based model architectures. But a number of generative AI vendors suggest that they can be done away with, more or less, through a technical approach called retrieval augmented generation, or RAG. Here’s how one vendor, Squirro, pitches it: At the core of the offering is the concept of Retrieval Augmented LLMs or Retrieval Augmented Generation (RAG) embedded in the solution … [our generative AI] is unique in its promise of zero hallucinations. …
Instead of relying on the LLM to take a company’s data and summarize it, or even retrieve it, the Voicebox LLM layer simply convertsRead More Source link
OpenAI has been hit with another complaint, after advocacy group NOYB accused it of failing to correct inaccurate information disseminated by its AI chatbot ChatGPT, potentially violating EU privacy regulations. According to Reuters, NOYB reported that the complainant in their case, a public figure, asked about his birthday through ChatGPT but received incorrect information repeatedly instead of being informed by the chatbot that it lacked the necessary data. The group also stated that the Microsoft-backed firm denied the complainant’s requests to correct or delete the data, claiming that data correction was not possible, and failed to provide any details regarding the data processed, its sources, or its recipients. NOYB reported that it had issued a complaint with the Austrian data protection authority, urging an inquiry into OpenAI’s data processing practices and the steps taken to guarantee the precision of personal data managed by the company’s expansive language models. Maartje de Graaf, NOYB data protection lawyer, said in a statement: “It’s clear that companies are currently unable to make chatbots like ChatGPT comply with EU law, …
OpenAI is facing another privacy complaint in the European Union. This one, which has been filed by privacy rights nonprofit noyb on behalf of an individual complainant, targets the inability of its AI chatbot ChatGPT to correct misinformation it generates about individuals. The tendency of GenAI tools to produce information that’s plain wrong has been well documented. But it also sets the technology on a collision course with the bloc’s General Data Protection Regulation (GDPR) — which governs how the personal data of regional users can be processed. Penalties for GDPR compliance failures can reach up to 4% of global annual turnover. Rather more importantly for a resource-rich giant like OpenAI: Data protection regulators can order changes to how information is processed, so GDPR enforcement could reshape how generative AI tools are able to operate in the EU. OpenAI was already forced to make some changes after an early intervention by Italy’s data protection authority, which briefly forced a local shut down of ChatGPT back in 2023. Now noyb is filing the latest GDPR complaint …
Hallucination, Barbenheimer and acabó were among the new and breakthrough words of 2023, according to an analysis of the Guardian archive. And do you have “rizz”? Dozens of new words appeared on the newspaper’s pages this year. And while the majority of these relate to people and placenames previously unrecorded on the news pages, some words were truly newly minted. Foremost in this list was the unlikely phenomenon that was Barbenheimer, a portmanteau of two films released on the same day but light years apart in terms of content (the films were Barbie and Oppenheimer in case you spent July under a particularly large rock). Barbieland – although not a new entry – also surged in 2023. “Rizz” – in the sense of charisma (the word is derived from the middle consonant of the word) – appeared on the Guardian pages for the first time this year, but went stellar when Tom Holland insisted he had none. “I have no rizz whatsoever. I have limited rizz,” the Spider-Man actordeclared, spawning seemingly endless memes. Other new …
Jessy Lanza’s third album, 2020’s All the Time, traded in suggestion. The Canadian producer let fly little wisps of desire – “want / you” – on the breeze of her off-kilter, neon-hued club music in the hope that they might be reciprocated. Love Hallucination changes mode: Lanza is no longer asking but demanding orgasms, devotion and boundaries, sometimes losing herself in the gulf between desire and reality. “So frantic with no purpose,” she sings over the wonky funk of Gossamer, sounding pleasurably lightheaded in pursuit of her needs. The artwork for Love Hallucination Lanza has credited this newfound boldness to her initially writing these songs for other artists. Yet Love Hallucination isn’t cosplay but an affirmation of Lanza’s unique ear. Her tactile heavy bass, cirrus-wisp synths and spun-sugar falsetto have deepened: the low end is diamond-hard, her playful freestyle-inspired melodies and moods glimmer like the light refracted through the gem. Don’t Leave Me Now is at once prowling and prismatic in its hi-NRG; the glimmering Drive is a study in liquid and solid. There’s pop …
