Hallucinating Bots: The High Price of AI's Misguided Confidence

The Gist

AI concerns. Over half of AI professionals question AI's future reliability.
Balancing act. Striving for a fine line between AI creativity and accuracy.
Staggering cost. Alphabet lost $100 billion in value due to AI misinformation.

In the age of artificial intelligence, there's a strange irony — machines with the ability to process vast amounts of data faster than any human could, sometimes get it wrong in high-stakes ways. Dubbed “hallucinations” in the AI community, these are confident responses that starkly deviate from an AI's training data.

A glaring illustration of this came from Google's chatbot, Bard, which confidently asserted a fact about the James Webb Space Telescope — a fact that swiftly led to a staggering $100 billion loss in market value for Alphabet.

In July, Microsoft launched AI-powered buying guides in Bing, but media outlets reported that Microsoft’s own promotional materials included hallucinations about the quality and specs of certain products.

As companies race to lead in AI capabilities, the challenges of balancing accuracy with AI creativity have become a defining concern, shaping conversations about the future of this transformative technology.

On Tuesday, Fortune reported that hallucination issues have many companies contemplating their future AI path, adding that “companies of all stripes now face a common dilemma: The competitive pressure not to fall behind in the A.I. race and the responsibility for the safety of the tools they release.”

In a recent survey, Databricks and Dataiku, found that of the 400 senior AI professionals who responded, more than half (55%) said that fears around AI are justified and that they’re more worried than excited about the future of AI.

“I probably trust the answers that come out of ChatGPT the least of anybody on earth,” OpenAI CEO Sam Altman recently told an audience gathered to see him during a stop in New Delhi.

But Altman also said he thinks they will get “the problem of hallucination to a much, much better place.”

“I think it will take us a year and a half, two years, something like that,” he said. “But at that point we won’t still talk about these. There’s a balance between creativity and perfect accuracy, and the model will need to learn when you want one or the other.”

In other AI news...

OpenAI Puts Its Money Where the Media Is: NYU's Ethical Journalism Drive Nets $395K

Recently, Yahoo Finance highlighted Google's efforts in creating a generative AI tool intended for drafting news articles. Although it's been presented to major publications like The New York Times and The Washington Post, the tool prompts crucial questions, like, should AI be incorporated in newsrooms — and if so, how can the public ascertain if the article they just read was written by a human or by AI? According to the outlet, among the major English-language newspapers, only one, The Financial Times has committed to ensuring its printed content is not penned by machines.

In response to some of the urgent challenges facing the journalism sector, including the rapid onset of AI, NYU's Carter Journalism Institute is introducing the Ethics and Journalism Initiative. This forward-thinking program will delve into pioneering research areas, highlighting the nuanced relationship between journalism and technology.

Fueling the initiative's ambitions, OpenAI has endorsed the project with a generous grant of $395,000.

“OpenAI is committed to ensuring that artificial intelligence is both beneficial and safe, and we welcome independent research, workshops, and discussions to help ensure a positive role for AI in the news industry,” Tom Rubin, chief of intellectual property and content at OpenAI, said in a statement. “We are excited about the potential of the new Ethics and Journalism Initiative and very pleased to support its goal of addressing a broad array of challenges journalists face when striving to practice their profession ethically and responsibly, especially those related to the implementation of AI.”

From Beta Beats to Global Grooves: Spotify's AI DJ Knows Your Tune

Since its beta launch in February, Spotify’s DJ has been a hot topic among social media users.

Billed as “a personalized AI guide that knows you and your music taste so well it can choose what to play for you,” the app was originally available only to users in the US, UK, Canada and Ireland.

But on Tuesday, Spotify announced that premium users in in Europe, Asia, Africa, Australia and New Zealand, can now also access the English version of the app. Available via the "music" tab in the Spotify mobile application, DJ tailors the listening experience for users with a handpicked selection of tunes, interspersed with spoken AI commentary along with playful chatter and song insights.

Aptos Labs Collaborates With Microsoft to Integrate AI and Blockchain

In a significant move to reduce the complexities of onboarding new users to web3, Aptos Labs has announced a partnership with tech giant Microsoft. The collaboration will leverage Microsoft's Azure OpenAI Service to expedite the global adoption of web3.

For many, the utility of blockchain remains obscured by its perceived complexity. Through this partnership, Aptos seeks to address those challenges by introducing the Aptos Assistant, an AI tool aimed at bridging the divide between the traditional internet (web2) and the emerging decentralized internet (web3). The assistant will help users, including developers, navigate the Aptos blockchain ecosystem, offering guidance on everything from creating wallets to developing decentralized applications.

“Artificial Intelligence and blockchain technologies are quickly converging for one important reason: they are both generational breakthroughs that profoundly impact the evolution of the internet and shape society,” Mo Shaikh, CEO of Aptos Labs, said in a statement. “Together with Microsoft, our shared vision is to ensure that this technology is accessible to more people and organizations than ever before.”

AI Tweet of the Week: An AI Archive From 1872

Stan Blade, the Dean of Agricultural, Life & Environmental Sciences (ALES) at the University of Alberta found a reference to artificial intelligence that dates back to 1872. Take a look.

One of the first works to address #ArtificialInteligence From the Samuel Butler novel #Erewhon (1872). #MachineLearning pic.twitter.com/8UjXis26p4
— Dean Stan Blade, PhD P.Ag. KSLA 🇨🇦 (@DeanALES_UofA) August 9, 2023