High-Performance Computing Revitalization Act of 2005

HIGH-PERFORMANCE COMPUTING REVITALIZATION ACT OF 2005 -- (House of Representatives - April 26, 2005)

Mr. SHERMAN. Mr. Speaker, I thank the gentleman for yielding me time.

Mr. Speaker, I believe I am the only one to come to this floor to oppose this bill. I will try to yield back at least a few minutes to the gentleman and hopefully they will be available should my comments result in comments of others that need rebuttal.

I support science. It is important to America. It is important to my district. But as we look at what we can accomplish, we also have to examine what we should try to accomplish. While we expand the tools of the human race, we must also look at the pitfalls.

This is an issue that I have been talking about for a long time. I first brought it to the floor 5 years ago and that is best illustrated by the fact that roughly 50,000 years ago was the last time that a new level of intelligence came to this planet. It was our ancestors, who said hello to Neanderthal, the only other intelligence on the planet that we were aware of at the time. It did not work out so well for the Neanderthal.

Today we are as a species looking at two exciting new technologies, each which is likely to create an entity, a life form, with a higher level of intelligence than human beings; and, in fact, a higher level by a differential that exceeds whatever differential there was between human beings and Neanderthals. One of these technologies is genetic engineering. And if this was a genetic engineering bill, I would not get to speak on it as long because there would be more members to speak against it, worried about the societal implications. But genetic engineering raises questions that should also be raised by computer engineering, because the kind of high-technology, high-performance computer which is the subject of the bill is an important step towards the development of an artificial life form that will exceed human intelligence.

We had hearings 2 years ago in the Committee on Science where the consensus of experts and I did not invite any of these experts, senior committee members did, (chiefly the chairman) they testified that we are roughly 25 years away from a computer that exceeds human intelligence.

Now, I do not know whether it is 25 years or whether it is a bit longer or a bit less, but should we go headlong into developing the next intelligent species on this planet without even including, in the slightest, in our legislation something to say ``let us examine whether this is something we want to do, and whether we want to have any controls.''

The truth is, Mr. Speaker, we do not know whether we are creating Data from ``Star Trek, The Next Generation,'' or whether we are creating Hal from ``2001: A Space Odyssey.'' We know that the future will look like science fiction. We just do not know which science fiction book or movie.

Last year when the Committee on Science considered this same bill as H.R. 4218, the gentleman from New York (Mr. Boehlert) and I reached an agreement on an amendment that would provide for looking at the societal implications of future advances in information technology. That amendment was included in the bill that passed this House. Specifically, it directed the National Science Foundation to support research into the implications of computers, both hardware and software, that were capable of mimicking human ability to learn to reason and to make decisions. Likewise, the nanotechnology bill which passed both houses, and is now law, provided for even more extensive review into the societal implications, including explicitly the implications of developing levels of intelligence that exceeds those of human beings. But H.R. 28 strips out the provisions that were included in prior legislation. This draft says we will do nothing to look at the societal, the ethical, the environmental implications of what we are doing, and we will rush headlong into trying to do it without the slightest thought of whether we should do it.

My amendment in committee was defeated 17-19 on what was unfortunately, and inexplicably a party-line vote. My amendment put forward just a few weeks ago was identical to the compromise language the chairman and I reached in the 108th Congress.

Now, the importance of understanding how artificial intelligence will be achieved through information technology, how it will impact society--that importance has not decreased since last year. The amendment should be included before this bill leaves this House.

Now, I know there are those who say it is okay to create a computer that exceeds human intelligence and that is self-aware because it will not have hands and will not be able to act except through human beings. Trust me, there are those amongst us who would sell hands to the devil for a good stock tip. If you create Pandora's box, it will be opened.

Now, H.R. 28 deals with the creation of high-performance computers; and as I said and want to say again, the testimony before our committee was that we are 25 years, and this is not one crackpot, this was a consensus; the range was 20 to 30 years between now and when we develop a computer that exceeds human intelligence. And it is not just me. The DARPA, the Defense Advanced Research Project Agency, has on its Web page the statement that its mission, supported by this bill, is to develop a computer which will ``learn from its experience, be aware of itself, and be able to reflect on its own behavior.''

So part of our government is engaged in trying to create maybe Hal, maybe Data, while here in the Congress we pretend that it is impossible, that it is not an issue worthy of reflection. DARPA is going to create a reflective computer, but we do not have a reflective Congress.

Now, I understand that H.R. 28 is an important bill to set goals and priorities in high-performance computer research development with a number of different agencies, including DARPA and its subsidiary agencies. What I do not understand is why there is such resistance to studying the implications of this research. We cannot and should not plunge ahead without a provision to study these implications.

Join me in rejecting this bill on suspension. A bill with this level of implications should not be considered under a suspension of the rules. Send this bill back to the Committee on Science. Have the Committee on Science create a balanced program. Overwhelmingly, this bill should deal with supporting the technology, marching forward, achieving all of the goals that the preceding speakers have indicated. But then let us also put in the bill just a little language to say that we ought to look at the implications: Whether it is likely that this technology will create an entity more intelligent than human beings? Whether that entity is likely to be self-aware? How we could either cause or prevent such self-awareness? What are the societal and ethical implications of having a slave entity reflective, intelligent, and commanded to do what we instruct, without so much as the minimum wage?

So let us pass this bill next month, after the Committee on Science can provide some balance to it.

BREAK IN TRASCRIPT

Mr. SHERMAN. Mr. Speaker, I thank the gentleman for his additional generosity. I wish to respond to the comments of the gentlewoman.

Mr. Speaker, I will submit for the RECORD the section of DARPA's mission statement that I referred to previously where DARPA itself indicates that its mission, using the funds provided by this Congress, is to create a computer that is self-aware and able to reflect on its own behavior.

Mr. Speaker, we have thrown around terms as to what is close and what is not. It just comes down to whether 25 years, 30 years, is something close enough for us to be concerned about, or should we be concerned about only the immediate future? I would point out that we are not going to have self-aware computers for at least 10, maybe 15 or 20 congressional elections. And so if that is how we measure time, self-aware computers are a long way away. But when we approve construction projects and roads, we do not build bridges that are going to collapse in 25 or 30 years, and we assume that human beings will be the only intelligent species using those bridges.

If we are concerned when we build infrastructure for things 20, 30, 50, 100 years down the road, then we should be even more concerned with this bill. And we should not pass this bill in this form and say, well, we will worry about these issues when they come up in some subsequent decade. In addition, it is put forward that we will just have the scientists and the research community figure out how to deal with these issues. That is perhaps the problem, because if we provide the support exclusively to the hardware and software scientists and nothing to those who will consider the societal implications, the ethical implications, the philosophical implications--then no one will be looking at those issues, then we will not have done our job to provide a balanced, scientific research bill. That is why I am voting ``no.''

The material I referred to previously is as follows:

DARPA Strategic Plan: Section 3.7: Cognitive Computing (Released February 2005)

Many elements of the information technology revolution that have vastly improved the effectiveness of the U.S. forces and transformed American society (e.g., time-sharing, personal computers, and the Internet) were given their impetus by J.C.R. Licklider, a visionary scientist at DARPA some 40 years ago. Licklider's vision was of people and computers working symbiotically. He envisioned computers seamlessly adapting to people as partners that would handle routine information processing tasks, thus freeing the people to focus on what they do best--think analytically and creatively--and greatly extend their cognitive powers. As we move to an increasingly network-centric military, the vision of intelligent, cooperative computing systems responsible for their own maintenance is more relevant than ever.

Despite the enormous progress in information technology over the years, information technology still falls well short of Licklider's vision. While computing systems are critical to U.S. national defense, they remain exceedingly complex, expensive to create, insecure, frequently incompatible, and prone to failure. And, they still require the user to adapt to them, rather than the other way around. Computers have grown ever faster, but they remain fundamentally unintelligent and difficult to use. Something dramatically different is needed.

In response, DARPA is revisiting Licklider's vision as its inspiration for the strategic thrust, ``Cognitive Computing.'' Cognitive computers can be thought of as systems that know what they're doing. Cognidtive computing systems ``reason'' about their environments (including other systems), their goals, and their own capabilities. They will ``learn'' both from experience and by being taught. They will be capable of natural interactions with users, and will be able to ``explain'' their reasoning in natural terms. They will be robust in the face of surprises and avoid the brittleness and fragility of expert systems.

The benefits from this cognitive computing thrust will be profound. The increasing complexity of military systems means that the level of expertise needed to maintain them is also increasing--as are the staffing requirements for virtually every military function that uses computing and communications technology. By creating systems that know what they are doing, and that can configure, maintain, and adapt themselves, we will be able to drastically reduce the staff needed for operations centers, forward command posts, and even in support of small dismounted units and special operations teams. Cognitive computing technology will also help us to deal with the increasing tempo of operations and the complexity of plans, such as Air Tasking Orders and joint hostage rescue operations plans, by allowing computers to tap into the accumulated knowledge of past experience on behalf of their human partners.

Along these lines, DARPA's Personalized Assistant that Learns (PAL) program will create intelligent personalized assistants for many tasks, such as a commander's assistant, an intelligence analyst's assistant, or a decision-maker's executive assistant. These assistants will interact with their human partners by accepting direct, naturally expressed guidance to learn their partner's preferences and procedures. Then, they will be able to anticipate the human's needs and prepare materials to be ready just in time for them. These new and unprecedented artificial helpers should reduce military staffing needs in many key places and will help ensure decisions are made in a timely fashion and with the best possible preparation.

To meet these challenges and seize these opportunities, DARPA has structured its work in cognitive computing to catalyze innovative work in single cognitive systems. collaborative teams of cognitive systems, and collective cognition from large numbers of small non-cognitive elements. Each area will demonstrate the power of merging reasoning, learning, perception, and communication technologies. These areas will be supported and complemented by broad-based technology efforts in the hardware, software, and integration techniques needed.

The strategic thrust of cognitive computing is a template shaping DARPA's core technology foundation work in information technology.

Mrs. BIGGERT. Mr. Speaker, I reserve the balance of my time.

BREAK IN TRANSCRIPT

http://thomas.loc.gov