When dropping out might be right choice for students

Many university faculty members would be disconcerted to have their doctoral students give up on their degrees, but Stanford emeritus professor of computer science Jeffrey Ullmann is nonplussed.

Having supervised 53 doctoral candidates at the California university, he was used to the occasional student disappearing – to join, or sometimes start, a new company.

When the ones who didn’t return include Google founder Sergey Brin, you can see why his take is a bit unconventional on the importance – or not – of finishing a degree.

“I’ve seen it both ways,” he says over lunch in Dublin, where he was visiting as part of his new role as chair of the international advisory board of the school of computing at the National College of Ireland, with overall responsibility for the strategy of its new cloud computing centre.

“Sergey went off to found the company, never got his degree, and I don’t believe he will ever miss it. On the other hand, look, if you join a company and it’s not successful you can come back and do your thesis.”

With a long roster of former students who went on to either found their own business or become influential in companies including IBM, Intel, Oracle and Microsoft, Ullmann is used to encouraging students to take such opportunities.

One, he told, “If the company is successful and you don’t join it, you’re gonna kick yourself, and kick me, forever for not letting you do that.” The company was successful and was bought by Amazon. The entrepreneur returned four years later to finish his PhD with Ullmann.

Another opinion

Despite its reputation as an entrepreneurial hothouse, Stanford provides little formal training in the area, he says.

“Basically, it’s a matter of the culture – it’s what you see around you. The students ahead of you go out and start a company or start working for a start-up and you say, ‘I could do that too’.” But he adds that Stanford, crucially, always valued producing software that people would actually use.

“Not so long ago I got an email from a student asking, can a PhD thesis be both useful and acceptable as a PhD thesis and I said, yeah, I think it can. He then said would I therefore go and fight with his adviser who was apparently of another opinion. Being useful was a proof that your thesis was not academic enough.” At Stanford, the opposite view would hold.

Ullman received his own degrees from Columbia University (BSci) and Princeton (PhD). Prior to joining Stanford in 1979, where he chaired the computer science department from 1990-1994, he was on the technical staff of Bell Laboratories, and on the faculty of Princeton.

“I think my secret throughout my career is that I don’t boss students. I sort of try to keep out of their way and I think that’s what Sergey needed, and a number of other students as well. Sergey didn’t really need an adviser,” he muses.

“I knew what they had was really, really great. They had a side-by-side demonstration; you could issue a query to their system, and to one of the others out there, and you could see how much better the response was to their system.”

Nonetheless, it wasn’t clear that Brin and Google co-founder Larry Page, also a Stanford postgraduate, had a search breakthrough that would prove the basis of a new company. “We had no idea that it would make a lot of money. In fact, they talked to Yahoo and apparently Yahoo was willing to give them a million dollars for the patents. But, they weren’t going to develop them, they were just going to sit on them. And these guys were really willing to , if Yahoo were willing to let them develop them.”

But Yahoo wasn’t – and the rest is history. Still, that choice was fraught with risk. “It sounded like a lot of money at the time,” recalls Ullmann, and neither Brin nor Page had any clear notion of how Google might generate revenue. “If you look at their original paper, there’s a footnote that says, ‘We do not believe advertising is a way to support search’,” laughs Ullmann.

“But, of course, what they were talking about was the advertising of the time, where it took 10 seconds to load some graphic or other and you couldn’t see the results until you saw that ad.”

What worked much better was their “snippet idea” of providing a fast, text-only advertisement. “I don’t know who had that idea, but they very quickly went to a snippet ad where they could deliver it faster than they could deliver the search query – and that worked.”

Ullmann, a seminal figure in the theory and application of computer science, well-known to generations of students for 16 textbooks, was adviser to Brin because of his expertise in the area of map-reduce functions – these are algorithms that allow large computing problems to be broken down into smaller pieces with a single end result.

Map-reduce works particularly well with large database questions, says Ullmann, and is very useful in working with increasingly common “big data” problems.

An original application for map-reduce was the “Page Rank” algorithm used by Google – its special way of evaluating millions of web pages, and determining the overall importance of each with what remains a regularly updated, and highly secretive, formula.

The map-reduce function allows big problems once limited to government-funded scientists working on expensive hardware to be solved by anyone using lots of inexpensive computers. Cue a new generation of entrepreneurs.

Free lunch

“For the scientific community, the government paid for their computing equipment, so they don’t care . They used the high-quality stuff. But if you were a Google or a Yahoo or a Facebook, you have to worry about what it costs because you’re paying for it yourself.”

He has maintained close ties to Google. Brin and Page placed him on the company’s technology advisory board at its founding. He’s recently started consulting for one of the groups at Google working on a cloud-based system. “But I just do it basically for the free lunches.”

That’s no doubt an understatement of his role. But he says he’s been retired for a decade “and basically I’ve sort of reached a stage in my life where I like to help out”. This is what also brought him to the NCI.

“I go pretty far back. When I started at this, we didn’t really even think of ourselves as computer scientists.”

After he worked at Bell Labs for a year, “the head of our department said, ‘I’ve just heard of this term computer science – so we’re going to call ourselves computer scientists’. And that’s how I became a computer scientist.”

The field was tiny. “I started out and did my thesis in coding theory, a kind of mathematical branch of engineering, and I got involved in automata theory mainly because I needed a summer job. The first courses I taught were automata theory; that led me into compiling, and algorithms in general. You could spend a month, and you were an expert in one of these fields.”

Except of course, he is an expert in those fields and others, with a string of awards and fellowships to his name.

His website includes some pithy opinion pieces, including a polemic on four-way stop signs at intersections in Palo Alto, and on speedbumps.

“Yeah, they drive me nuts.”