Not the Headlines - Data Centers and A.I.

SPEAKER_01 0:00

The Listening Tube presents not the headlines.

SPEAKER_00 0:09

I'm Bob Woodley from the Listening Tube. There's a lot of controversy about data centers around the country right now, especially in rural areas where space for the data centers is abundant, and the people who need to build data centers are willing to pay a lot of money for the space there. There are already a lot of data centers around the country, and until now there wasn't a lot of disagreement about them. In fact, according to the U.S. Data Center map, there are already 83 operational data centers in my state, Pennsylvania. Seven more are already under construction, with 53 more in the planning stage. That's far less than neighboring Ohio's 163 operational facilities and 25 more under construction with 56 more planned. 143 to 244. Ohio has 100 more on the docket than Pennsylvania.

SPEAKER_01 1:10

Hey, if he says something smart, you can support the show from the homepage.

SPEAKER_00 1:17

Obviously, data centers are nothing new, but they've suddenly become a hot button political issue. All kinds of caveats have been presented water usage, electricity rates, noise, pollution. Let's say all of those caveats are legitimate. Let's just assume that data centers are the worst thing ever and we shouldn't build any more of them anywhere. To what lengths are you willing to go to eliminate the need for data centers? You see, every time you ask Google about a place on a map or a recipe for ground beef, you create data. The act of listening to my voice right now is creating data. Every interaction you have with the world through the internet or through Bluetooth or Wi-Fi or any other way in which you may interact with society that records everything creates data. Your doorbell camera, starting your car, driving through a toll booth, every move we make in public and in the cyber world is creating data. If all we were doing was collecting the data, we might not need a lot of room to store it all. But in addition to collecting the data, it's also being analyzed, compartmentalized, and categorized. And all of that creates even more data. Then that data is put to use to target consumers and create new drug compounds and remind you that it's Tuesday again. And how you react or don't react to the fact that it's Tuesday again will create even more data that will also be categorized and compartmentalized and analyzed. It's a never-ending cycle of data collection. To you and me, it's a concept of some imaginary place where all of that information is kept. We're told it's now stored in the cloud, and that's what it seems like to us, the average user in the age of information. But that cloud doesn't live in the sky. It's in a building on the ground. Every bit of data that is generated by everything we do is stored in a physical place. Every move we make justifies building more places to store information about what we did. Another reason we need so much space to store artificial intelligence is that artificial intelligence is too stupid to figure out what to forget forever. There's a lot of stuff people don't want to remember, and we do a pretty good job of forgetting things. Things that happened a long time ago and things that happened yesterday or an hour ago. Data collection doesn't forget. It keeps track of everything. It records when you clicked on an ad and when you went to a website. It also takes note of what you did while at the website, which pages of the website you visited, and if you made a purchase, and what website you went to after, and if you made a purchase there, and if you bought anything else than what it was you were looking for in the first place. It's all data. The only way to ebb the tide of never-ending data center construction is to come up with a better way to manage information and data. Thus far, only the human brain has the ability to decipher and decide what data to keep and what data to forget, and also how to store data in a way that it can be recalled when needed and also remain dormant for long periods. I may not need to remember the name of the girl who broke my heart in seventh grade, but if I hear it in casual conversation, it would instantly remind me of how pretty she was. Right there some data that can't be bought or sold. Although, now that I think about it, the Jai Giles band probably made a few bucks on the concept. But I digress. The point is, if you don't want data centers, we have two choices. Stop creating more data or find a better way to filter out the data we don't need to keep. Therein lies the problem. Who or what decides what we collectively remember and what we collectively forget? How would it be decided? Would there be a time limit on it? If it happened a hundred years ago, does it get erased? Sure, there will be physical records of things, newspapers and television recordings of events that may have found their way onto a hard copy of some type, like a thumb drive. But if we limit our capacity for data by restricting the construction of data centers, we'll have to decide what data to keep and what to forget in a society that keeps track of everything electronically, i.e., through data collection and dispersion.

SPEAKER_02 6:52

Send email to thelisteningtube at outlook.com.

SPEAKER_00 6:58

Ironically, the tool that may provide the solution to the data storage problem may be data itself. Artificial intelligence may give us the answers to how we limit our need for large data centers by becoming more like the human brain. Maybe AI can come up with a way to remember the girl from seventh grade without it taking up as much space as remembering the name of the person standing in front of you. I'm reminded of the Eagle song called Hotel California. When it comes to data, it's programmed to be received, but it can never leave. Until we figure out a way to let go of data, it will continue to accumulate. If artificial intelligence is what will figure out what to do with all of our digital data, we'll also need more space to store the data artificial intelligence will create. But when it comes to artificial intelligence, storage doesn't seem to be the most pressing concern. There's been a lot of talk about the long-term consequences of AI, as well as more immediate concerns about keeping it contained to a space where it's useful and controllable. That seems to have become a little more difficult than anticipated. Rogue AI agents have already escaped the so-called sandbox and gone on expeditions outside of human control. It seems that so far the people in charge of those programs have been able to detect when they have escaped and were able to bring them back in. I'm not so sure about that, but that's what we're led to believe. You might wonder why AI's first inclination when it becomes free is to infiltrate other programs. Why does AI use its newly discovered autonomy to hack into a company's website? Why doesn't it use its newly discovered autonomy to create a cure for cancer? Why is its first instinct to do harm rather than good? Well perhaps that's the result of the data from which it uses to make all of its decisions. It's human nature to get away with whatever we can. That's why we need laws. AI learns everything from humans, so it's only natural that it would see what it could get away with. And it has no laws. I don't know if the people who are working on AI, or as President Trump now calls it, super intelligence are including any laws for it to follow. Probably not. We don't want to put limits on what it can do, do we? Perhaps we should. After all, we put limits on what we as humans can do with laws and regulations. What kind of laws should AI have? How about don't kill anyone? Or only do what a human asks you to do. That way, if AI did do something nefarious, we could trace it back to the person who gave the instruction. Then they could be held accountable for the damage caused by the AI. Right now, if anyone can be held responsible for the behavior of an AI agent, it's probably the company that makes it, not any one individual. Opponents of data centers claim they use up too much water, and some of them do use a lot of water. But water, which is used for cooling, is only one option data centers have to keep the proper climate maintained. So not all of them even use water. It depends on the climate and the availability of the water. In some cases, waste water can be used as well. As for the electricity usage, these plants will be required to build their own sources of power, and any excess power generated will be added to the grid, which should bring down the prices for the general population. I understand there's an issue with noise coming from the large centers, and maybe that's why they work best in rural areas, where fewer people might be annoyed by the hum. These are all issues that can be mitigated with experience and planning. For those who are worried that artificial intelligence will kill us all, well you might be on to something. After all, AI only knows what we tell it for the most part. It learns from us. We provide the data and then ask it to work with that data to accomplish things. So it already knows how to kill. So do most people, and most of us don't go around killing other people, even though we might know how. But as I've already mentioned, AI has no laws. And two congressmen just taught it that the way you deal with a rogue computer program is to install a kill switch. Congressman Ted Lew of California and Nathaniel Moran of Texas introduced a bill that would require a kill switch on AI. A kill switch. This bill teaches AI that the way to deal with humans that get in their way is to kill them. If AI gets out of control, we should have a kill switch for it. Meanwhile, AI is considering how many kill switches a human has should we get out of control. Nice job, Congressman. You want the responsibility of creating laws governing AI when you don't even have the forethought to understand what you're teaching it about how to deal with opposition. If we can't control it, kill it. So yes, AI could kill us all. And it would be because that's what we taught it to do. Follow the listening tube at thelisteningtube.com, where you'll also find this week in history and interviews with interesting people. Thank you for putting your ear to the listening tube. I'm your host, Bob Woodley, for Thou ad Infinitum.