Interview with George Sadowsky, Internet Pioneer and Hall-of-Famer

Although I never had an earth-shaking epiphany regarding information technologies, there were several realizations that were important to me in understanding the evolution and role of digital technology. 

I started in computing in 1957, and my first few jobs included using and running mainframes that operated in batch mode. When I had to replace one such computer and saw what was possible with emerging interactive (time sharing) computing models—in my case a Digital Equipment PDP-10. I saw the future of computing as interactive, and I eventually persuaded my management to buy a PDP-10 to replace a batch computing environment.

In 1978 I was working at the United Nations, and we were putting the first computers into African countries for statistics and population census projects. The target countries had an inadequate technical base, and so we were limited to providing computers that came with their own support systems, such as customer engineers, power conditioning equipment, and so on; so therefore mainframes. This was a very expensive proposition. Microcomputers were just coming out, some of which had small but serious processing capability. I realized that microcomputers solved both our cost and repair issues (buy redundant machines, ship back for service), and I initiated an experiment to do population censuses in microcomputers—first in Cape Verde, then in the Comoros Islands and Mauritania, and over time in larger countries. The experiments were successful, and for me the future of the mainframe was limited.

Finally, in 1980, while helping to install the first computer in Rwanda, I had the occasion to visit the university library. It was contained in one room, and the technical books occupied about one linear meter of shelf space. I remember that both the concept and the label of “information poverty” hit me at the same time. Responses to address technical problems were sparse or non-existent in the country, and even the knowledge that they could be found elsewhere did not exist. Further, with no effective communication with the outside world, needed resources could not easily be located, and even if they were, the time and money cost of acquiring them might be prohibitive. 

This realization of ubiquitous information poverty in most of the developing world of made a deep impression on me at the time, and was the motivation that led me and a group of similarly motivated colleagues to start the Internet Society’s network training courses for people from developing countries which took place during 1993-2001. We taught people how to link their countries to the global Internet, how to build and manage routed IP networks, how to search and find information on the Internet, and how to evolve and manage incipient national networks.

Well, I was an atypical graduate student in that I had had five years of industrial experience before starting graduate work in economics. New management in Kennedy’s Treasury Department was somewhat aware of what computers could do and was willing to experiment. The older civil service staff members that had been using manual methods for producing revenue and distributional impacts of tax proposals were suspicious but not hostile. The project, which appeared serendipitously, was a major turning point in my career.

I was fascinated by Joe Weizenbaum’s Eliza program when it was written in the 1960s. It has been resurrected, and you can now access it online.

The program implemented a Rogerian therapy model, the basis of which is to reflect the patient’s thoughts and feelings in a way that delves deeper into problem areas and makes associations that could be new to the patient. So the model was moderately easy to implement. I was impressed with the fact that after Elize became public, there were reports of people using it in a secretive manner, indicating that it had appeal to real people who were apparently willing to commit private thoughts and feelings to it. Perhaps Eliza should be recognized as the mother or grandmother of existing bots oriented to therapeutic purposes.

It really hasn’t changed very much. Of course what is considered artificial intelligence at any point in time depends very much on what the state of knowledge was at the time. The science fiction writer Arthur Clarke once wrote, “Any sufficiently advanced technology is indistinguishable from magic,” which is to say that technology much more advanced than what we understand will seem like magic to us. A payroll program in the 1890s would have passed Clarke’s test and been dubbed “AI”.

As a profession, we are strongly attracted to the currently new bright and shiny object, which happens to be AI large language models at the moment. Now this attraction is a good thing because for every new tool, method or model, we want to explore what it can do for us, but often the newness and shininess fascinate us and lead us to predictions that don’t materialize. Does anyone remember Second Life and the Metaverse? LLMs have the capacity to do some things well. But if you understand that every single sentence generated by an LLM is a hallucination, and that making models better means reducing but not eliminating the possibility of residual hallucinations, then some limitations of these models become apparent. 

For many observers of the AI scene, the power of having the computer speak your language has created some leaps into fantasy. In addition, the current focus on LLMs ignores that there are other approaches to what we call AI, and advances are being made in other areas that may turn out to be equally if not more important than LLMs.

My bottom line is that I expect to see useful tools emerge that are based on LLMs, but that the jump to AGI is far from happening.

I was a member of the Internet Society Board during 1996-2000 when the future of the domain name system was being seriously explored. It was increasingly clear that some significant expansion of the name space was desirable, and of the various proposal ICANN was selected. Depending upon a private organization for such a fundamental element of communications infrastructure was unusual, and led to the multistakeholder model of governance that evolved for ICANN. 

In 2004 I was asked by Vint Cerf to head the ICANN Nominating Committee, which is really a director selection committee. You don’t easily say no to Vint, so I agreed to chair that committee from 2005-2007 and got very interested in ICANN issues both from a technical and an economics perspective. My interest in being part of building this new industry grow, so I ran for the Board and was a member for 2009-2018.

Those years saw the introduction of over a thousand new top level domains to the domain name root. A lot of time was spent figuring out how the rules should be interpreted given the desire for security and stability of the system. The issue of plurals was (and still is) vexing; for example, should ‘hotel’ and ‘hotels’ both be allowed? How much user confusion would be associated with such choices? What should be considered a geographic name and therefore protected? Ask the proponents of .spa and the residents of the town of Spa in Belgium. Ask Amazon and ask the countries that are part of the Amazon river basin in South America; you’ll get very different answers, What about religiously related domains, such as .kosher and .halal? 

Another major issue was the restructuring of the governance system to accommodate a satisfactory withdrawal from United States government contractual control of core ICANN processes? The Board and the community worked separately, not without considerable friction to reform a governance structure that the community distrusted. The U.S. government had set a deadline date of September 30, 2016, and in addition there would be a national election in November with uncertain results.

Finally, there were fights to claim specific domains considered by the registry community to be highly marketable. The contest regarding legally correct ownership of .web, initially delegated in 2016, was challenged and is still tied up in adversarial proceedings. (Disclosure: After I left the ICANN Board, I was an expert witness in some of those proceedings.)

Where do I start?!

In terms of work, whatever your current job is, learn as much as you can from it and do it well. Unleash your curiosity and follow where it leads you. Don’t be afraid to try new things, but learn to differentiate between acceptable and unacceptable risk. Be loyal to your organization, but beware of excessive and blind loyalty. You have entered a field that moves quickly; maintain your technical edge in both breadth and depth by spending sufficient time regularly being involved in something new. 

In terms of career, choose a path that provides some benefit to humanity. Identify people who can best mentor you and ask them for their guidance. Take advantage of serendipity. As you progress in life, share what you know with and teach those who are following behind you. Be generous in helping others to succeed.