Episode Transcript
[00:00:00] Elizabeth Gearhart: There was news that rogue AI agents were going to destroy humanity.
[00:00:05] Arshavir Blackwell: If you were working on a bomb that could destroy the entire planet, you wouldn't keep working on it and say, but we'll work on it just a little bit slower.
[00:00:13] Richard Gearhart: Do you think that artificial intelligence will ever achieve consciousness?
[00:00:18] Arshavir Blackwell: It's my belief that if you do have a system that is quote unquote conscious, and we still don't really know how to define that, it'll be something with some kind of a biological component mixed in.
[00:00:31] Richard Gearhart: In we're kind of going full circle. We created these silicone chips to do the processing and now we're incorporating biological matter.
[00:00:41] Elizabeth Gearhart: This is real AI Use Cases Business Owners Roundtable. Here's how real companies are using AI right now.
This is real AI Use cases Business Owners Roundtable. I'm Elizabeth Gerhart.
[00:00:55] Richard Gearhart: And I'm Richard Gearhart and we're your hosts.
[00:00:58] Elizabeth Gearhart: We're here today with Arshavir Backwell. He has been in AI practically his whole life and what was the precursor to AI? And if you want to find out more about what he's doing and see his amazing company, go to subscribe.insidetheblackbox AI on substack. Welcome, Arshavir.
[00:01:17] Arshavir Blackwell: Hi. I'm very glad to be here. Thank you.
[00:01:20] Elizabeth Gearhart: You're welcome. So I want to dive right into a controversial subject.
[00:01:25] Arshavir Blackwell: Okay.
[00:01:25] Elizabeth Gearhart: Since you're an AI expert, this. We've been asking people how they use AI in their business. We could do that too. But recently, last week. Was it Richard? Yeah. There was news that rogue AI agents were going to destroy humanity.
[00:01:39] Richard Gearhart: Yeah.
[00:01:41] Elizabeth Gearhart: What do you think of that?
[00:01:43] Arshavir Blackwell: Not too worried about that.
You know, for one thing, I point out that we have a lot of other much more pressing concerns in the world right now. Things that I'm way more worried about than artificial intelligence. And in fact that some of those concerns, I think AI may help find the solutions or the partial solutions to them. I mean, there's a strong degree of hype and anti hype when it comes to artificial intelligence and there always has been because it's just such a cool kind of science fictiony Star Trek technology that it lends itself very well to both predictions that we're all going to end up having so much free time, all we'll do is play tennis all day and robots will cook our meals for us, which isn't happening yet. And the other side, it'll be like the Terminator and take everything over and destroy it.
The few examples of what people are calling rogue AI are really just the Systems not having the proper protections to keep them, they're able to go and manipulate other computer systems. It is true, you do need to have guardrails for that.
But I don't think that they're going to be able to tap into NORAD for, for example, and launch a missile or something like that, which is the image that I see people hinting at when they talk about rogue AI or drones being controlled by AI and coming and bombing us. Now, people could do that. People could launch a whole drone attack on some population. And they do. But I'm not really concerned about AI doing it. And if you like, we can explore. There's a lot of reasons why that. That's just not a very realistic scenario.
[00:03:45] Richard Gearhart: So let's go into those reasons. I get what you're saying. I do think that there's a lot of walking into the haunted house around Halloween and stuff going on where people just like to make up stuff and get everybody all riled up, including themselves.
But it is a new technology and I don't really know that anyone knows what the boundaries are.
But I would be very interested in understanding why you feel that the threat is not so real.
[00:04:16] Arshavir Blackwell: Well, for one thing, what you've got to keep in mind is who is saying this. And a lot of the people that are saying this are the big corporations. And it's not necessarily the case that they have some hidden insight into what these systems are capable of doing. It's also a form of positive hype because what's coming across to people is, wow, this is super, incredible, powerful technology.
I better invest in it. And I think that's.
[00:04:45] Elizabeth Gearhart: I want to just stop you there for a sec because I was talking to my coach this morning and she was saying that this has brought AI front and center to many, many more people.
[00:04:55] Arshavir Blackwell: Yes, right, exactly what's happened. And I feel it's like with publicity, the whole. The old saw, just as long as they spell my name right, that's all I want.
And so in other words, no bad publicity is bad publicity. There's no such thing. Right.
So what I think is going on here is exactly what you've said, is that these people have kind of hit on. And you know, the reasons for that are, for one thing, like if Anthropic, for example, or OpenAI really believed that this was going to stop humanity, that it was all over, then they would stop what they were doing right away. If you were working on a bomb that could destroy the entire planet, you want to keep working on it and say, but we'll work on it just a little bit slower.
You would.
[00:05:43] Elizabeth Gearhart: Because they live here too, don't they? They're still human, right?
[00:05:46] Arshavir Blackwell: Exactly. So I know that the really wealthy ones have bunkers, but there's a lot of them involved that don't have bunkers.
So I think that the way they're reacting to their own news, I think is a very strong clue to the fact that there's really a very strong element of hype. It's not to say that one of these couldn't get loose, but, you know, we've always had trouble with viruses and other kinds of rogue agents on the Internet. I remember way back in the 80s, there was a big Internet virus. Everyone has forgotten about this, of course, that someone had let loose on the. I don't even know if it was called the Internet back then, it was so long ago. But they had let it loose on the network and it had required some amount of effort to bring it back down. And so this. I don't think a lot of what people are talking about is even that new in terms of computation.
And what you have to remember is in the background, there's always a person controlling these things still. And they're really not as magically autonomous as people seem to believe. I mean, just go in there and use one for a while and you'll see it's an amazingly powerful tool, and it can do a lot, but sometimes it makes just really stupid mistakes as well.
And it's kind of like if you've seen videos of humanoid robots, they do all these incredible things, and then they start flipping out and falling to the floor and shaking or something like that.
And so those are the things that I'm not particularly worried about taking over right now. There was a movie, actually, that had exactly this scenario as its premise. Not the Terminator, an older movie called the Forbin Project.
And there, a computer basically blackmails the entire world with nuclear weapons. But the thing is, they gave the computer access to the nuclear weapons. I always like to point out we haven't done anything like that, and I don't think we're likely. I mean, there's still. People never know.
[00:07:55] Richard Gearhart: Well, yeah, it may happen.
[00:07:58] Arshavir Blackwell: The thing is, if you look at the structure of the way these systems are designed and the way our nuclear arsenal design all of this stuff, there's just no way for one thing, a data center, for example, you could just unplug it and they all have kill switches, they all have places where the power supply is routed through, and it would be trivial to just turn those off.
And actually a friend of mine wrote an article about just this fact and he points out that the problem with the doom saying is it does not take into account how really very fragile these systems are. A strong burst of electromagnetic energy, all of them are fried. These are systems that people have to spend an enormous amount of time keeping working.
Right. You always hear about uptime and these large server centers and there's all of this staff who, you know, they wake up at two in the morning because something's down.
So these are not systems that if humans come kind of walked away from and just let them do what they wanted to do, they would not last on their own. They would not function on their own for very long even if we did nothing to stop them. And that's why I'm not so Archavir,
[00:09:15] Richard Gearhart: do you think that artificial intelligence will ever achieve consciousness?
Because that's really, I think what we're talking about here.
[00:09:25] Elizabeth Gearhart: Singularity they call it.
[00:09:26] Richard Gearhart: Yeah, singularity.
So it can think and act independently.
[00:09:32] Arshavir Blackwell: I would argue that thinking and acting independently is possible, but that it's different from consciousness. And the problem is consciousness. I do have a degree in cognitive science, so we've spent a lot of years seminars talking about what is consciousness. And the problem is that no one can really satisfactorily define, even in human beings, what, what exactly consciousness is. Because if something is alert and responsive to you and seems to have a memory and all of those other kinds of things, that does not make it per se conscious. It does not make it something that has awareness and feelings and sensation and that knows what it is, what's going on in the world.
The current models are really just really, really, really complicated computers processing really, really complicated mathematics at lightning speeds.
[00:10:35] Elizabeth Gearhart: Can I ask you a question about that, Arshavir?
[00:10:38] Arshavir Blackwell: Yes.
[00:10:38] Elizabeth Gearhart: Okay. I just wanted to ask a quick question.
Is it true that it's still binary code running everything?
[00:10:45] Arshavir Blackwell: Yes.
[00:10:47] Elizabeth Gearhart: So keep going. Sorry.
[00:10:48] Arshavir Blackwell: Yeah, keep. So you have just very, very powerful systems, but they're running in parallel.
A lot of mathematical computations. And at the end of the day when you see it respond to something that you've said or something that you've asked or, you know, whatever, given it a picture. That's all that's going on.
So the problem with consciousness, it's not that I don't think, think that some kind of a synthetic system couldn't have consciousness.
I just don't have a good. No one has a good definition. No one has a good explanation for what consciousness even really Is I will tell you one thing. If you want to get worried about something, and I'm not worried about it myself, but something that's a little bit on the sci fi side.
There are now data centers. I forget. I believe it's in Taiwan. There's now somewhere in Southeast Asia, there's a data center where they have 18 million human neurons that are in a system and are part of the process of doing the information processing. And so those are biological systems.
And it's my belief that if you do have a system that is, quote, unquote conscious, and we still don't really know how to define that, it'll be something with some kind of a biological component mixed in.
So it'll be something with human or something similar neurons.
And then on top of that, maybe this synthetic computer component as well, with millions and millions of chips and transistors. But I think they culture them. I don't think they'll take a personal brand.
[00:12:39] Richard Gearhart: Well, what role do the neurons play in the system?
[00:12:43] Arshavir Blackwell: I don't know much about that system, but I would assume basically what's happening is it's doing just the same kinds of things that the servers in the data centers are doing, which is you're training it when it gets some sort of input to provide some kind of output, which is just exactly what the neural network paradigm is. Now, you know, by the way, to your question about binary. Now, neurons are not really binary.
Neurons have an analog quality to them. So those might actually be an example of systems that aren't binary. And I wish I knew more about exactly. I don't know how much of the details they've really released about what task they're working on or how they're hooked up to the rest of the system, but it is, I think, pretty remarkable.
[00:13:32] Elizabeth Gearhart: It blows my mind that the AI we have today is all built off ones and zeros. Right? In the very early days, I read they would unplug it and plug it back in. So unplug would be off and plugging it back in would be on. That would be your 0 and 1.
That's amazing to me. I still can't really wrap my neurons around that.
[00:13:51] Richard Gearhart: I think it's interesting that we're kind of going full circle. We created these silicone chips to do the processing, and now we're incorporating biological matters, I guess, with the hope that it can become quicker and more efficient than the chips.
[00:14:08] Elizabeth Gearhart: So I do want to switch gears a little bit. Arshavir. This is supposed to be about how we're using AI in business. But I did want to address the elephant in the room, which was this fear mongering terror for maybe good reason, maybe not. Anyway, but I know you have a business where you help people develop their own LLMs that are private.
What do you feel is the most powerful way to use the current AI that we have right now in business to get a real ROI to really get financial benefit?
[00:14:37] Arshavir Blackwell: What you're talking about, of course, are frontier models like ChatGPT and Claude and all the various incarnations of those that seem to come out with a new update every week. And I think that for someone just getting into using AI, which is most people getting familiar with how the system works and runs, just going, and if you don't have an account, making an account, it's 20 bucks a month and going in and using it for as many day to day real AI type tasks as you can.
And by real, what I mean is a lot of people still use it just for things like search.
So instead of typing a question in to a search engine which are becoming increasingly powered by AI as well, they just type it in Claude or ChatGPT. And that's really just the very surface of what these systems are able to do. So for example, if you have a meeting that you've recorded, what I do is I put all of my meeting transcripts into CLAUDE and it's able to synthesize all of that information across all of the different meetings. So that if I have a question about a certain topic that spans a whole bunch of different meetings, or if I have a meeting with somebody and I'm not quite clear on what they meant by something, I can ask the AI and it will usually be able to provide a very good sort of explanation of what's going on. Because when we're in these conversations, we only have so much bandwidth to process what's being said. And what this allows you to do is to go back over the meeting or the call that you've had with a fresh set of eyes so that you're not the only one who's figuring out what the details of this meeting are or what the details of the problem stated in there are, there's a lot of uses that one can find. So for example, every time I go to the doctor now or every time I get a medical report, I put it into claude.
And CLAUDE now has a very in depth record of all of my medical visits for the past year or so. And that includes all the tests. So I can give it tests that they do and it'll tell me in plain English what those tests mean, but it also is able to correlate all the other stuff that that's in there. So I'm not a doctor, so I'm not going to know that this thing from this report over here on one day and this thing from this report over here on another day. Those two things may have some information when put together that I would never recognize and that a doctor might never recognize because you've seen two different doctors and they're very busy and they're not going to be able to look out through your entire medical history.
So those right there are just two really powerful examples of what I think people should start doing with these systems.
Not just for the value that they add in your workflow and in your day to day life, but also because they get you much more comfortable with using them in other contexts as well.
[00:18:16] Elizabeth Gearhart: Which one do you think they should start with? I think chat's probably the most popular.
[00:18:21] Arshavir Blackwell: That's a great question. That's a very personal question. I used to use Chat GPT a lot more.
I prefer Claude. Now, I can't really give you a cut and dry reason why I prefer Claude more. I think that in terms of my financial outlay, because it can get very expensive, the $20 is a flat rate for a relatively small number of tokens. Now, most beginning users are not going to go above that, but someone that's using it all the time is going to spend a lot more on tokens.
So I think that Claude maybe is slightly a better value. I'm sure there are people that would disagree with that. There just are subtle differences between the two systems that are hard to really put into words. And I think that you have to really one try the other and see which you feel is giving you the best kind of answers for the kind of work that you're doing. For instance, one problem I have with both of them is they give you these really long, annoying answers. You can just say, hey, how do I do X?
And then there's a wall of text that comes back at you and you're like, oh my God, I don't want to read all this. I just wanted a simple answer to my question.
And you can actually steer them by saying, no, give me a short answer. Give me, give me an answer in simple English. Don't take more than a paragraph to respond and put it in easy to understand language. And it will do that. So both systems, you can mold them.
[00:19:53] Elizabeth Gearhart: Okay, I want to go to Richard for just a sec. What is your favorite one right now. Are you using different ones for different.
[00:20:00] Richard Gearhart: I use Copilot because it's within the Microsoft ecosystem and confidentiality is important.
I don't think the results are as good as ChatGPT or Claude. I use ChatGPT for anything that's really not confidential or client related, mostly because I've just been using it for a long time and it seems to really know me well. I've been told that if you want to switch LLMs that you simply ask, you know, if I wanted to switch to Claude, I'd ask Claude to create a prompt for me to download all the information from ChatGPT and put that into ChatGPT.
And then ChatGPT would answer and I could copy and paste all the information that Claude thinks is most important and put it into Claude and I could start using Claude, but I haven't done that yet.
[00:20:56] Elizabeth Gearhart: So I'm using four of them right now. I hadn't been using Perplexity for a while, but Perplexity's kind of gotten cool because it has agents that go out and do your queries. So Richard gave me this intense query to do and I put it in Perplexity and it gave me the spreadsheet that he wanted. Chat couldn't do it.
I love Chat for images. Right now. Chat is doing great images.
[00:21:19] Arshavir Blackwell: Yeah, I think it's very good for images. I think that's true.
[00:21:22] Elizabeth Gearhart: And I love Gemini for Gemini gives you the most tokens, but I love it for multimodal. So we do a lot with video, obviously. And so Gemini is great for video. But also when we're analyzing our YouTube stats and our everything, I always ask Gemini because Google owns YouTube and so I figure I'm going to get the best answer. But I do sometimes spot check it. So I use ChatGPT, I use Gemini, I use Perplexity and I do use Claude. But I guess the way my brain works and the kind of things I'm doing, I don't really get great answers from Claude. Maybe I'm not giving it the right prompts. Yeah. So for writing, for trying to write content, I think ChatGPT is friendlier.
But I do for facts especially I use all of them because Claude is a lot more factual, I think. I don't think it hallucinates as much as the other ones. Maybe.
[00:22:16] Richard Gearhart: Hard to know.
[00:22:17] Elizabeth Gearhart: Hard to know though. That's why you gotta check them all against each other.
[00:22:21] Richard Gearhart: Speaking of hallucinations and also, for example, relying on AI for medical information, I do think that it's a great tool for that because I do that with ChatGPT. I go to a doctor and I put my test results in and it gives me very credible sounding answers to my questions. And as an attorney, as an intellectual property attorney, we have clients now that are using CHAT or LLMs to write us emails asking us questions about certain topics. And the challenge is that you may get a five page email that's going to take half an hour to respond to and you're not really sure the client needs like every single detail that the LLM is asking for.
And in the legal profession, we have to charge for our time. That's how we pay our bills.
And so we've gotten to the point now where we're just asking the clients, here's a couple lines that we think you need to know about if you want us to answer the rest of the questions. We're happy to do this, but we have to bill you for the time. And the point I'm trying to make is the LLMs sound so authoritative and so it creates the impression that, that they're right and sometimes they're not always appropriate for the situation. Which is what you were saying about the long answers. Right.
[00:23:47] Announcer: You have been listening to Real AI Use Cases Business Owners Roundtable with hosts Elizabeth Gearhart, marketing expert and PhD researcher using AI every day, and Richard Gearhart, entrepreneur, business owner and intellectual property attorney specializing in innovation. We we hope you found this valuable. Join us again for more stories. Because the future of business is driven by AI.
