Mozilla sets up UK data firm
A "third way" to sell data to AI labs.
Good morning, it's time for me to leave the beach and hide some Easter eggs. This newsletter is a big believer in public holidays and will be back on Tuesday. Wishing you all a Happy Easter break and fruitful hunting until then.
A message from Locai Labs
SPONSORED: The UK has missed the boat on foundational AI right? Not so fast. Meet one entrepreneurial family that thinks otherwise. London startup Locai Labs has developed a UK LLM by post-training an open source model using just one node of GPUs and a tiny budget. How? Why? Find out more and get in touch with them or just chat to it at GB1.ai.
News In Brief 🩳
Clifford's new chapter: VC Matt Clifford is stepping back from some of his work at Entrepreneurs First to focus on "ways to contribute to UK tech sovereignty", he said yesterday. The former government AI advisor will become the startup accelerator's non-executive chairman, while continuing as chair of ARIA.
Apply to the frontier: ARIA is looking for partner organisations to help it get cutting-edge AI capabilities into the hands of UK researchers. They want partners focused on frontier AI for Science tools and autonomous labs. Applications close May 21.
No thanks: NHS staff are boycotting Palantir's Federated Data Platform over ethical concerns, the FT reports. Labour MP Chi Onwurah, meanwhile, who supports the government breaking its deal with Palantir, said she was driven by concerns around the contract rather than ideology against the company.
Cluster cash: UKRI is giving regions up to £20m to support innovation clusters, it announced this morning. Areas getting the money include defence-tech in south-west England, critical minerals in south Wales and autonomous vehicles in the Oxford-Cambridge corridor. DSIT also said yesterday that Professor Leszek Borysiewicz is its preferred candidate to be the next chair of UKRI.
Opinions please: The Competition and Markets Authority has opened a call for evidence to get views on changes proposed by Google and Apple to their app stores, following the watchdog's digital markets investigations.
Mozilla sets up UK data firm
Scoop: Mozilla is launching a UK company today to help organisations sell their data to AI developers. The Mozilla Data Collective gives developers access to hundreds of clean datasets, while the data provider gets control of where their data is used plus payment.
Bad diet: Its founder EM Lewis-Jong said: “This generation of AI was fed a diet of data scraped from the internet, and it shows. We need a tech future that is more multilingual, multicultural and multimodal – basically, more human.” They said “indiscriminate scraping” was not the way to get there and organisations needed control of how their datasets are used.
Enter Mozilla Data Collective: “Our goal is to build a real, pro-people alternative," they said. "You should be able to talk to the person behind the dataset, see the metadata, the provenance."
Why now? Model developers are searching for better ways to access high-quality data, but Mozilla, the Californian non-profit best known for creating Firefox, believes there is a lack of infrastructure for fair data sharing, which hinders AI’s development. “We want builders and creators to work and imagine together," EM said. "When you break this mission down to first principles, a data collective is absolutely core to enabling that."
Third way UK: Mozilla has set up the firm in the UK because it reckons the country is well-poised to create a “third way” between what is usually presented as a binary choice between respecting data ownership or winning the innovation race, EM said.
Strong start: The social enterprise has signed up 160 organisations to make around 850 datasets in more than 300 languages available. Data ranges from Afghan literature to Nigerian folktales and are being used by thousands of public labs, journalists, researchers and tech companies, including UK startups, EM said.
Original Sin: "I think one of the original sins in tech is when a product’s impact ambition, its growth strategy, and its revenue strategy don’t align," EM said. "With Mozilla Data Collective, we’ve built an aligned model: we grow the platform’s curation of community datasets, which drives more people to build tech for those communities, and generates revenue for them and for us."
How it works: It makes money by charging the downloader, but not the data uploader, a five percent fee. The group was first announced in November, but today is setting up as a UK company. It is backed by multi-million-dollar seed funding and is the first social enterprise incubated by the Mozilla Foundation.
'Government can turbocharge'
This week's Q&A is with Anne-Marie Imafidon, chair of the government's Women in Tech Taskforce. She co-founded Stemettes in 2013 to encourage more young women into STEM careers.
1) What do you hope the Women in Tech Taskforce can achieve that other initiatives have failed to?
The Taskforce is systemic and backed by a Secretary of State (Liz Kendall) who is truly about change in the sector, which drives better outcomes for all of society. Ultimately I hope we’ll achieve a better culture, better retention and more funding for women working in tech - from academics to founders, entry-level to leaders, employees and the self-employed.
2) How much is it down to government to push industry to increase opportunities for all and how much should it be down to businesses themselves?
It’s a balance. Both have a role to play: this is why the taskforce being run by DSIT is a great opportunity. Business has only done so much thus far: proper government intervention can turbocharge what more needs to be done.
3) In your career and work at Stemettes what are the repeated challenges you encounter with getting more women and girls into STEM?
Having an industry that fosters belonging (rather than hostility towards them), feeling a proper sense of connection and community when entering into the field, proper awareness of the plethora of pathways into and within the field, and having women and girls see where their identity matches up with driving all that’s possible technically. These are the systemic barriers that we encounter again and again.
4) What are your hopes and fears for how AI may impact society?
My fear is that those creating and driving AI decisions do so without due regard for the socio-technical nature of its impact. My hope is that the benefits of AI are felt by all and well distributed due to smart, wise and equitable considerations of the possibilities brought about by such technology.
5) Give us a piece of advice you’d tell your teenage self.
When you go together, you go further. Look for a crew, community (I wish I had Stemettes) to celebrate with, commiserate with and learn alongside.
Happy holidays, back Tuesday.
Tom