Create a free account, or log in

Neural Notes: ACCC probes LLM market, Meta’s new Llama, and KFC’s drive-through experiment

In this edition: Australia’s competition watchdog reveals a new focus on large langage models (LLMs); Meta touts its “most capable” models yet; and KFC causes a clucking flurry at the drive-through.
David Adams
David Adams
AI accc
ACCC chair Gina Cass-Gottlieb. Source: SmartCOmpany via AAP Image/ Lukas Coch

Welcome back to Neural Notes, the weekly newsletter where Tegan Jones tackles the latest developments in artificial intelligence. I am writing today’s newsletter, but rest assured Tegan will return next week.

In this edition: Australia’s competition watchdog reveals a new focus on large language models (LLMs); Meta touts its “most capable” models yet; and KFC causes a clucking flurry at the drive-through.

ACCC competition callout

For its tenth and final report in the Digital Platform Services inquiry, the Australian Competition and Consumer Commission (ACCC) is considering the ties between LLMs and constrained market competition.

Unlike prior reports from the inquiry, which asked if AI is skewing competition in online search results, this one will investigate the barriers for entry in the LLM sector itself.

An ACCC issues paper, published Thursday, reflects several potential competition concerns, like:

  • high upfront and ongoing costs of maintaining an LLM blocking out new entrants,
  • the creation of ‘feedback loops’, where entrenched players are able to extract and process data at a scale impossible for newer entrants,
  • anti-competitive self-preferencing, tying, and data access restrictions taking root in LLM platforms,
  • the potential for LLMs to orchestrate cartel-like behaviours without human intervention.

“Adoption has been extensive, and this technology continues to expand and develop at a rapid pace,” ACCC chair Gina Cass-Gottlieb said in a statement.

“Generative AI products and services may present new opportunities, but also new challenges with major implications for our work.”

The stakes are high here. Australian regulators undoubtedly want local startups to compete with big global players in a fair and meaningful way.

But the fear of over-regulation among AI advocates is very real, suggesting this report, and its outcomes, will be watched quite closely.

Stakeholders are invited to share their views on generative AI with the ACCC by August 23.

AI use a key concern in Nine Publishing strike

Unionised journalists, sub-editors, and production staff employed by Nine Publishing — covering The Sydney Morning Herald, The Age, The Australian Financial Review, and online sites Brisbane Times and Watoday — have launched a five-day strike from this morning.

The Media, Entertainment & Arts Alliance (MEAA), the union representing media workers, confirmed the strike will go ahead after members voted down a last-minute enterprising bargaining offer put forward by management.

The revised proposal — offering pay bumps of 3.5%, 4%, and 3%, over three years — would still be insufficient to help workers keep up with rising CPI, said MEAA Media acting director Michelle Rae.

The union said management could not guarantee it would pause layoffs, after already announcing between 70 and 90 redundancies in its publishing division.

Tellingly, the use of AI in the newsroom is also a sticking point.

Unionised staff are “taking a stand” for “ethical and transparent use of Artificial Intelligence,” Rae said.

Nine Entertainment, parent company of Nine Publishing, has reportedly vouched for its AI safety practices in communications to staff.

In a memo obtained by AdNews earlier this month, the company said “all our decisions” regarding AI usage “will be guided by our values.”

“Audiences trust us because they trust our people,” the memo continued.

“These principles will continue to evolve as the technology and business models associated with it continue to evolve.”

The strike comes just ten days after MEAA launched a campaign urging the federal government to impose new opt-out rules for generative AI, in an attempt to shield the work of journalists and creatives being used to train AI.

Meet Meta’s latest Llama

It was just a few short years ago that Meta pinned its hopes on the metaverse, and its vision of a ‘decentralised’ social internet.

These days, Mark Zuckerberg has legs. And Meta has seemingly channeled its energies into developing open-sourced LLMs it says are capable of challenging the liked of OpenAI’s GPT-4 and Anthropic’s Claude 3.5 Sonnet.

Meta’s Llama 3.1 405B is its latest effort, a model it says is capable of use cases like “long-form text summarisation, multilingual conversational agents, and coding assistants.”

The company is also touting its ability to play nice with third-party platforms, giving developers more ways to build out a real-world product.

Crucially, it is open-source, a model Zuckerberg himself said is preferential to the closed shop model offered by competitors.

“I believe that open source is necessary for a positive AI future,” Zuckerberg said in a Tuesday blog post.

“Open source will ensure that more people around the world have access to the benefits and opportunities of AI, that power isn’t concentrated in the hands of a small number of companies, and that the technology can be deployed more evenly and safely across society,” he continued.

Is open-sourced AI really more competition-friendly than the alternatives? Maybe that’s one for the ACCC to consider next.

Law Society of NSW, LexisNexis team up on AI glossary

Rapid developments in machine learning and LLMs, combined with the sector’s growing cachet among non-technical observers, mean the layman is increasingly exposed to complex terminology.

To legal professionals keep up to date with those terms and their meaning, the Law Society of NSW has partnered with legal-tech giant LexisNexis on a new AI Glossary.

Penned and maintained by the Law Society of NSW’s AI Taskforce, the resource will help our learned friends differentiate their RAG from their NLP.

In a statement, Law Society of NSW President Brett McGrath said the glossary “will be an indispensable resource for practitioners using AI tools and platforms, and for those seeking understanding about this burgeoning technology.”

Not bad timing, given AI’s encroachment on the legal sector itself. 

The glossary is available via the Law Society of NSW’s website.

KFC tests automated voice-chat for drive-through orders

KFC has become the latest fast-food business to dabble with automated voice ordering systems.

7NEWS reports the fried chicken chain is testing a system allowing drive-through customers to place an order without directly speaking to restaurant staff.

The trial is purportedly active across five outlets: South Penrith, Wetherill Park, Minto, Mt Druitt, and Frenchs Forest.

Crucially, for customers who want their Zingers to come with a touch of personality, the system will be optional to use.

Staff numbers will reportedly go unchanged through the trial, too.

While automated ordering systems are one of the most obvious use cases for AI in the restaurant sector, it has not proved an instant success.

McDonald’s last month ceased its AI-powered ordering trial in the United States, after customers recounted bizarre errors on social media.

The system, delivered in partnership with IBM, may not be the last from the Golden Arches: the brand said it would likely revisit AI technology in the future.

This enables our latest models to support advanced use cases, such as long-form text summarization, multilingual conversational agents, and coding assistants

Canva says (AI-powered) creativity is good for growth

In a bruising market for many businesses and startups, Canva is positioning creativity — and AI-powered creativity — as not just an aesthetic nicety, but a necessity.

Coinciding with the release of new research from Harvard Business Review, suggesting more creative ‘leaders’ are using AI tools compared to their followers and laggards, co-founder and chief product officer Cameron Adams put it this way:

In a business world focused on the bottom line, it’s easy to lose sight of the value of creativity. The findings highlight that creativity isn’t just a complement to business growth, it’s foundational to driving long-term success. Ultimately, innovation and creativity is what will differentiate the leaders from the laggards; it’s what drives growth in a challenging landscape.

Of course, Canva’s entire business model is relies on the idea of accessible creativity, and the ability to enterprises to deploy that creativity at scale.

Yet Adams’ most recent comments make the link explicit in a very interesting way, aligning Canva more closely with accounting and workplace management SaaS than a graphic design platform.

Other AI news this week

  • Workplace safety technology powered by artificial intelligence will have an outing at the Workplace Health & Safety Show, held in Sydney on October 23 and 24 this year. Its Knowledge hub, sponsored by ecoPortal, will raise awareness of next-generation tech integrations. Think CCTV systems with AI-powered plugins, capable of detecting ‘near misses’ on site, or AI-powered drones to help assess risks in building zones. “AI, put to good use, will foster a culture of continuous improvement in safety practices and employee wellbeing, supporting safer, healthier workplaces,” said Helene Seidel-Sterzik, ecoPortal co-owner and CEO. 
  • You are probably already familiar with the idea of AI chatbot as a quasi-romantic companion — at least, you’ve likely seen one of the dozens of Hollywood movies featuring a digital sweetheart. In those big-screen depictions, it’s usually a man pining after his ones-and-zeroes boo. Yet new women are seemingly increasing their use of AI-powered romantic chatbots, as per Axios.