University of California, San Diego

What it is

The University of California, San Diego is a public research university in the La Jolla neighbourhood of San Diego, California. It was established on 18 November 1960.^[1] Its place in the history of artificial intelligence comes from two separate periods: researchers on its faculty were among those who made neural networks a serious subject of study again, and from 2017 the university built data science and computing into institutions of their own.^[2][3][6][7]

Neural network research

David Rumelhart earned a PhD in mathematical psychology at Stanford University in 1967 and joined the UC San Diego faculty in the fall of 1974. He stayed until 1987, when he returned to Stanford.^[2] In 1986, with James McClelland and the PDP Research Group, he published the two-volume *Parallel Distributed Processing*, which presented connectionist — that is, neural-network — models to a wide audience.^[5][2] The connectionist idea is that knowledge need not be stored as explicit rules; it can sit in the pattern of activity across many simple units, with learning adjusting the strengths of the connections between them.

An October 1986 paper in *Nature*, "Learning representations by back-propagating errors", lists Rumelhart and Ronald J. Williams at UC San Diego's Institute for Cognitive Science, together with Geoffrey Hinton of Carnegie Mellon University.^[3] Hinton had been a postdoctoral researcher at UC San Diego before moving to Carnegie Mellon.^[4] The method named in the title works by computing the error at a network's output and sending it backwards through the layers, so that connections far from the output are adjusted as well as the ones beside it. The second half of the title, learning representations, points at what this buys: the features a network uses are not supplied in advance but formed during training.

Data science and computing

In 2017 UC San Diego announced a $75 million gift for data science from alumnus Taner Halıcıoğlu,^[6] and in that year established the Halıcıoğlu Data Science Institute.^[7] The institute was officially launched in March 2018,^[6] and was dedicated on 2 March 2018 in the auditorium of the San Diego Supercomputer Center, at an event led by Chancellor Pradeep K. Khosla.^[6] Its senior staff and faculty were housed in the East Building of the San Diego Supercomputer Center.^[6]

The university's own account places the founding of the data science institute in 2018 and the founding of its School of Computing, Information and Data Sciences in 2024.^[1] In July 2024 the UC Board of Regents first approved the school that now unites the institute with the San Diego Supercomputer Center.^[7] In April 2026 the university announced $125 million in support from Halıcıoğlu, an alumnus, for the Halıcıoğlu School of Data Science and Computing, which brings the data science institute and the supercomputer center together.^[7]

How the pieces fit

The two strands show the same university appearing in the field's history in different ways. In the earlier period the contribution took the form of papers and a research group: a faculty member, a postdoctoral researcher and their collaborators working on how networks can learn. From 2017 the contribution took institutional form — a gift, an institute, a launch, and eventually a school that also absorbs a supercomputing centre.^[1][6][7]

The dates are easy to mix up, and the reason is concrete. The gift was announced in 2017, the institute was established in 2017, it was launched and dedicated in March 2018, the regents first approved the combined school in July 2024, and further support was announced in April 2026.^[6][7] Different years, largely the same organisation at successive stages.