Nothing artificial about intelligence at world-first AI university
The Mohamed bin Zayed University of Artificial Intelligence, known as MBZUAI, is recruiting from leading AI countries and universities. “The plan is to attract the best and brightest from around the world,” Baldwin, formerly of the University of Melbourne in Australia, told University World News.
Though still small – the university has 52 teaching faculty and 202 students – it is leveraging “incredible density in the AI space, which puts us on a par with the top universities in the world in this narrow area”.
MBZUAI says that for 2022 it is among the top 25 universities globally in the Computer Science Rankings – CSRankings – in AI and the three specialist areas of computer vision, machine learning and natural language processing, as well as web and information retrieval.
“The number of papers presented per scientist at top conferences in the past three years puts its quality research output above rival countries such as Switzerland, Israel and France,” says the university. “Its faculty chaired five of the most prestigious global AI conferences in 2022.”
A national AI vision
Established in 2019 and licensed in 2020, the university is named after Sheikh Mohamed bin Zayed Al Nahyan, known by his initials MBZ, the third president of the United Arab Emirates. He is also ruler of Abu Dhabi, capital of the UAE and its second most populous city after Dubai, with a population of around 1.6 million people.
According to Timothy Baldwin: “The university has a vital role to play in many of the UAE government’s strategic objectives, with AI identified as a critical component for future growth and prosperity. MBZUAI’s strategic vision and mission works in parallel to position Abu Dhabi as a hub for the international AI community.”
UAE leaders created MBZUAI to educate and develop talent, foster an innovation ecosystem, and act as a strategic think tank for the public and private sectors, the university says. More specifically, it was founded to support the UAE National Strategy for Artificial Intelligence 2031 – a strategy to diversify and decarbonise the economy.
AI is one of the major planks of the future economy, with a target of 13% of Abu Dhabi’s GDP being AI generated by 2031. This could be achieved, among other ways, by AI enabled efficiencies in government or direct revenue from AI start-ups, foreign direct investment in the AI space or AI multipliers for local industry.
“It’s a very ambitious, very aggressive target,” says Baldwin. The university is seen as a catalyst to help the UAE achieve AI 2031 in three primary ways.
First, MBZUAI’s pipeline of masters and PhD students could be a prime contributor to UAE’s capacity in AI. “We provide graduates with as many reasons to stay as possible, in terms of linkages with government and industry, internships and programmes.”
A second contributor will be research, says Baldwin: “So, creating an AI ecosystem directly via the university and having the IP [intellectual property] that is generated, whether that is directly spun out into start-up companies or generates revenue sources in terms of licensing the IP.”
The third way is for the university to be “a magnet for AI investment into the region. Not just homespun AI start-ups, but also attracting AI companies from around the world, or non-AI companies that are trying to develop AI capabilities.”
MBZUAI has evolved and grown rapidly during its three years of operation. Initially it had two departments: computer vision, which encompasses AI that has anything to do with graphics, for instance images or video – face detection is an example of an application – and machine learning, which is core AI in terms of building models for data types, such as financial or logistics data.
When Baldwin joined MBZUAI a year ago, he founded the third department, natural language processing – the interfaces between AI and human language. ChatGPT is a good example, though only part of what the field is about.
The university offers research masters and PhD programmes in each of the three sub-fields. Its 202 postgraduate students include 161 pursuing masters and 41 studying for PhDs. The first graduating class in 2022 included 52 students.
This year, MBZUAI is launching four new departments. They will be in core computer science, robotics, statistics and data science, and human-computer interaction. “Moving forward, we’re talking about doing something in the health space as well as the business space, details to be determined,” according to Baldwin. Some 30 additional faculty will be recruited this year.
With the new departments, the university will add more applied AI, “but still building off that core and identifying AI adjacent components of those areas to develop a faculty base around.”
Attracting the AI world’s best and brightest is possible because of the attractiveness of the university’s critical mass of expertise, which is extremely rare. “I can walk down the hallway and I’ve got an expert in vision transformers, and another door down I’ve got an expert in causal inference.
“Having absolute world leaders in the three core areas of AI makes it a very attractive place to be. So we’re finding it easier and easier to attract top talent – partly as our name gets out there more, but also partly because of the unique environment that we’re hiring people into. It was a very deliberate choice, of course, to build up from that core of AI,” says Baldwin.
The university’s leadership is among those world leaders. MBZUAI’s president is Professor Eric Xing, a renowned scientist in machine learning, computational biology and statistical methodology, who studied at Tsinghua University in China before obtaining PhDs at both Rutgers University and the University of California, Berkeley in the United States.
Before MBZUAI, Xing was in the US as a professor at Carnegie Mellon University. He was visiting associate professor at Stanford, has authored or contributed to more than 400 cited research papers, and has been an associate editor of numerous US journals. Xing also founded and runs Petuum Inc, a 2018 World Economic Forum Technology Pioneer company that builds AI software for industry.
For his part, before the University of Melbourne, Timothy Baldwin was a senior researcher at Stanford University in the US, and had visiting positions at numerous institutions including the University of Cambridge in the United Kingdom, the University of Washington in the US, the University of Tokyo in Japan and Saarland University in Germany.
Prominent researchers on the MBZUAI board include Professor Anil K Jain, a distinguished computer scientist at Michigan State University; Professor Daniela Rus, director of the Computer Science and Artificial Intelligence Laboratory at the Massachusetts Institute of Technology; and Sir Michael Brady, emeritus professor of oncological imaging at the University of Oxford.
A diverse student population
The same international flavour applies to students, who are recruited globally – there are international students from nearly 40 countries. “We have a strong base of local students, as part of the university’s mission is to develop AI capabilities amongst the Emirati population and do AI talent development,” Baldwin told University World News.
In the last month or so, there have been recruitment drives in Sri Lanka, India, Germany, Switzerland, the US, Ireland and South Korea. There are both faculty-led and university-led recruitment efforts. Baldwin expects student numbers to reach around 500 within two years.
“It is very attractive for faculty to have such a diverse student population. Almost a third are female, which we’re very proud of. Very few top AI universities would come close,” he says.
“The melting pot of cultures is something that our students often comment on, in terms of a place to spend a period of time and be welcomed into a very diverse community, which makes it easier when moving to a new country.”
The UAE and China are the two countries most highly represented among students. The number of Chinese students has grown progressively over the three student intakes so far. But both countries are still minorities in the total student pool. Most academics come from the world’s top AI nations, such as China, the US, Canada, France and Australia.
The MBZUAI Executive Programme
As part of its mission to support the AI targets and development of the UAE, the university’s academics and administrators do not only deliver learning to students. The MBZUAI Executive Programme, launched in 2021, reaches government, industry and civil society leaders with top class AI orientation. So far, 120 decision makers in UAE have participated in three courses.
The 12-week intensive course, with hybrid bi-weekly sessions, was developed to support the need for AI savvy leaders across all sectors, says the university. It is delivered by world leaders in research and innovation in AI, across academia and business.
“It’s a remarkable programme. The speakers come and give first-hand insights into the development of different fields, the challenges, the complexities and where the field is headed – very much from a 10,000 feet perspective rather than from a nuts and bolts technological or mathematical perspective,” Baldwin says.
“Because it is targeted at government and business leaders, there’s a strong focus on the societal impact and ethical implications of AI, as a function of where AI is headed in the short-term future at least, what are the complexities likely to be as we head along that path?”
The idea of the programme is to help the UAE elite unlock the potential of AI to achieve smart management, increased efficiencies and enhanced productivity. “It will also promote greater engagement on AI between government, business and the scientific community to accelerate the development of future industries in the UAE while supporting the nation’s long-term strategy to become a global AI leader,” the university says in a statement.
Roles of AI – Where to from here?
Of course, AI is not new – “It has been sitting in your pocket in your mobile phone and in lots of smart devices for 10 years or so,” Baldwin points out. “It is in spam filters in your inbox, for example, and helps organise your photos on Facebook.”
But today the impacts of AI are becoming more tangible. “We will see more clear instances of AI, such as ChatGPT, where there is standalone AI technology that you interact with directly. People are getting very excited about the potential applications.”
Working on key areas of AI that need improving is what scientists at MBZUAI engage with on an almost daily basis, in terms of liaising with local industry and government and international partners, Baldwin says.
“There is a myriad of different application areas of AI. Despite what departmental structure might suggest, there is a lot of overlap between them.” The university’s critical density of AI expertise makes it very well geared up to work across the different aspects of AI.
Indeed, AI is a field with considerable interdisciplinarity. While there are different clusters of academics focusing on different topics, there is also a lot of crossover between them.
An example is a cross-department large-scale project relating to sustainable AI that the university is putting together as part of the Year of Sustainability in the UAE and in the lead up to this year’s COP28 United Nations Climate Change Conference.
“A model like ChatGPT costs about US$50 million to train and a huge amount of energy,” says Baldwin. “The trend in the field over time has been for bigger and bigger models leaving larger and larger carbon footprints, and more and more hardware and faster turnover in the hardware.”
MBZUAI scientists across sub-disciplines are looking at whether it is possible to buck this trend.
“Can we have equally expressive models with similar capabilities that we can train in much more efficient ways to considerably reduce the carbon footprint and make the models much more accessible to areas of the world that don’t have access to the high end, very expensive hardware needed to train these models at the moment?”
Further, a lot of faculty members have collaborations with people in fields completely outside AI. “AI is something that you apply to different problem types or different data types,” says Baldwin, who himself has collaborations in areas ranging from veterinary science and environmental sciences to pro bono legal assistance and scientific publishing.
There are a lot of interfaces with other organisations and other areas of expertise that are outside AI. “Building external linkages and deploying models over real-world problems to make change for the good largely means that you collaborate with a lot of external people in other areas. By definition, there is a lot of external interdisciplinarity beyond AI.”
ChatGPT is a new step in evolutionary AI change. “There have been truly impressive advances, but there are still a lot of things that technologies like ChatGPT are not very good at. It is very good at sounding convincing, but factuality is not a strong suit, nor verifiability. I would suggest it is more of a plaything than a viable tool for most contexts at the moment.”
Unsurprisingly, given that MBZUAI is a research university specialising in AI, it is less affected than a conventional university by ChatGPT and other evolving technologies. It does not need the general purpose capabilities that a model like ChatGPT is geared to deal with, and ChatGPT cannot provide insights into the sorts of assessments MBZUAI has.
“First, the AI field is moving so quickly that the data for ChatGPT tends to be stale. Second, because it is research, by definition there’s a lot of creativity that needs distillation – a lot of knowledge that needs to happen,” Baldwin says.
“If anything, AI expert students are encouraged to play around with AI technologies, as well as develop them themselves, of course. Whether any of these tools can be used as part of the research process or built into assessments in different ways, is very much a work in progress.”