Artificial but not so intelligent
After months of mainstream chatter around the technology, the Financial Times this week offered up “the sceptical case on generative AI.”
With the world’s biggest companies such as Microsoft and Google parent Alphabet throwing their weight behind the technology and declaring “that AI is the new electricity or fire,” “never knowingly outhyped” venture capitalists are also throwing their cash in pursuit of use cases for the emergent technology.
“Fifty of the most promising generative AI start-ups, identified by CB Insights, have raised more than $19bn in funding since 2019,” Thornhill writes. “Of these, 11 now count as unicorns with valuations above $1bn.”
Add to that predictions like those from McKinsey, which suggests AI will add between $2.6tn and $4.4tn of economic value annually, and it’s clear why the technology has begun to take over the world.
“But what if they are wrong?” Thornhill asks. Technologist Gary Marcus has already suggested the market may see a “massive, gut-wrenching correction” in valuations, “as investors realise generative AI does not work very well and lacks killer business applications.”
Marcus is co-founder of the Center for the Advancement of Trustworthy AI, and has long been a skeptic around the real “intelligence” of AI models.
He points to large language models’ propensity to “hallucinate”, that is to say, make things up, as one of their key drawbacks currently being ignored by the investors so keen to make the technology work.
While engineers work to correct them, Marcus suggests these hallucinations will remain a part of generative AI models and be unfixable using the current technology.
“There is a fantasy that if you add more data it will work,” he told Thornhill. “But you cannot succeed in crushing the problem with data.”
Thornhill also details how training future AI models on now polluted data sets (containing inaccurate information due to the meddling of current models) will lead to a greater incidence of hallucination and contribute to the “enshittification” of the internet.
Still, Marcus sets out a number of the real-world use cases for AI, even in spite of its less-than-perfect accuracy.
Whether those uses are enough to drive profits for investors – or whether the AI market is a bubble soon to burst – remains to be seen.
Would you bet on ESPN?
Legal Sports Report (LSR) this week brought us its analysis on Penn Entertainment’s decision to ditch Barstool Sports in favour of ESPN.
The agreement to launch ESPN Bet, LSR said, represents “the most coveted branding partnership in US sports betting.”
However, the deal will not be all sunshine and roses, as Penn now occupies the unenviable position of building a customer base for an entirely new betting brand in an extremely competitive market dominated by just a few giants.
The article harks back to The Stars Group’s deal with Fox Sports a few years ago, with a helpful reminder that “it did not work.”
The same fate, it argues, awaited other media brands attempting to enter the betting sector, including Yahoo, Sports Illustrated, Maxim and Fubo.
Not to mention, it added, Penn has already attempted some variation of the same strategy with its Barstool Sportsbook.“Even as the best performer of this group, Barstool Sportsbook’s national share is languishing in the low single digits in every market it serves. Its overall slice of revenue across its 16 active states is less than 5% and trending down as it prepares to join the graveyard of US sports betting brands.”
That reality makes it “pretty easy to make the case for why ESPN Bet will not succeed,” LSR suggests.
On the flipside, though, many argue that the current US betting market dynamics are not fixed “and perhaps even ripe for serious disruption.”
As the number one sports media brand in the country, boasting more than 100 million monthly unique visitors and 25 million subscribers to its ESPN+ product, “if there is any brand that can do it right now, it is ESPN,” the piece suggests.
The brand’s potential to become a betting behemoth largely depends on ESPN’s willingness to engage with the sector, the article adds, as well as the ever-important matter of product quality.
Penn has already said it does not expect ESPN Bet to turn a profit over the next 12-18 months, so there is a long way to go before the partnership has a chance to bare fruit.
Still, this deal probably has a greater potential to shake up the US sector than any other we have seen this year.
Research funders Down Under
Over to Australia now, as The Guardian expressed its concerns this week about a new research centre at the University of Sydney.
The Centre of Excellence in Gambling Research was launched this week at the world-leading university, but is funded with hundreds of thousands of dollars provided by stakeholders in the gambling industry itself.
Funding was provided by the International Center for Responsible Gambling (an American nonprofit group that funds scientific research on gambling addiction and is itself bankrolled by a veritable who’s who of American gambling firms), together with Flutter-owned Sportsbet, and Entain.
While the university “has been upfront about its links to the gambling industry,” and insisted that gambling firms would not be given the opportunity to “constrain or edit the research in any way,” many have expressed concerns about the industry’s involvement.
The relationship between gambling firms and the university is “troubling in part because it normalises the relationship with research institutions and the gambling industry,” said one commentator.
Another suggested that “For [gambling companies] to bolster and window-dress their social licence by trumpeting the Sydney university association hurts all Australians.”
Meanwhile, the head of the new research centre, Professor Sally Gainsbury, said: “This partnership is unprecedented and will allow us to translate research findings into effective real world, evidence-informed strategies to prevent and reduce gambling-related harms.
“Having access to major gambling operators is essential as it means we can conduct live trials and test the efficacy of interventions designed to encourage positive behavioural change.”
Indeed, as well as providing the funding for the research, the gambling firms involved will also supply the centre with de-indentified data on gambling behaviours, allowing researchers to better understand the impact of various interventions.
“This is a significant development in gambling research,” Gainsbury concluded.
While concerns around the origins of funding may be justified, perhaps it is only fair that the sector pays its way with regards to research into gambling behaviours.
Companies could surely be accused of not doing enough had they not supplied the funding, so this could well be a classic case of “damned if you do, damned if you don’t.”