What I want to do, talking about I have a view. This optimistically may surprise you about why I'm excited by this and how I think I think about it might be helpful you have come to this to understand these things as the universal personal interest. That's relationship that I think COVID-19 might be a useful framework to think about what these things are. Spoiler, good to the ends that you're not going to
come back to that.
Relax you're not going to lose your job today. So what is this thing? The stuff we're talking about? It's
we don't know.
We don't know whether this
your name smart mostly what we think of AI is all the things that it can't do yet. Right. So when we think of AI is all the things that it hasn't done yet that sort of
the stuff that it can do.
goes under different names, things like machine learning and
refining models, neural networks,
but that's actually the name of this device. Certainly talking about projectors is what it produces our Hollywood
entertainment movies.
That's the product and machines are kind of diseases, the names of machines. We only had a good name for what it is that produces and something starts to happen. I think we're all talking about it. It's Headlines, news, and it's something
that's also been the most
quickest adopted technology ever. Like from zero to 100 million people using homeless within days, months, weeks. It's incredibly satisfying. And so that's something I'm always thinking of asking how many people here are have used it at least once, I think it is.
Very obvious, there's a lot going on. And the question is what, what is what's special about why so so that's why so soon? And I think the answer is that there's a little story though. A famous song in America, John Henry, who was working on the railroads were the head of the nail in space and the Americans and there were some issues coming up until that time was something that humans did. And the thinking at the time among the people who were doing it was that this is something that machines could not do. Because there's too much dexterity too much. Actually needed to actually do this and the machines are no good about it. And the story, of course, is that I'm going to die trying to outdo the machine, and she kind of won and that's what we understand now is that, you know, the best and most mechanical jobs in rural areas, but this is not something that elevated human only job in the slightest. And I think that's what's happening right now is that isn't that AI is really shocking us with their abilities this and we're coming to realize that a lot of the things that we were doing, actually hurting mechanical, much more mechanical than we thought. So, you know, there was a list of things like playing chess, we thought when the time there's always something that humans to do and our turns that was in question are mechanical process than thought. Searching a billion pages of information was what librarians to do. And that might not turn out to be a very mechanical thing we could automate recognizing a face was seen one fire was something that was
particularly difficult for us we
have actually fell in capital due teaching chain. So we put down this list all these things are right now, one of the things is I was painting a picture must be very highly. Human life typically realized no, no, actually, it's pretty mechanical. You can actually synthesize and have a conversation which is the latest to pass desk work to your job. And so there is something going on and I like to use the word artificial smartness. To talk about it to get away from the AI is clear that many things in our lives already smarter than we are like our new calculator. way smarter you are interested to your navigation device, the GPS, your phone in one way or a student smarter. Mapping navigation and algorithm chatbots are way better as far
as when you are inside of things.
And so, you have these we have these these elements of smartness. And
the issue is for us is that
we have incredibly poor understanding our own intelligence. You know, how we think and that individual and collective. And so it's hard to think of this as it is like the festivals where there's just one dimension because higher higher kind of started to Klaus monkey. Us and so that's going to be wrong I don't know so far is smart. This is a very complicated mixture, many different kinds of modes, and of the various strengths and they of course, differ from person to person. But we can examples of animals where, in some cases, some of their abilities can actually see
what we can do in certain directions and many others, they don't because again,
they make some different types of fishing in there. And what we're going to do now with the AIS is that we're going to engineer them with a different complex of different kinds of things and some other dimensions will be higher than ours and many of them will be lower because there's always a trade off. That's the engineering maximum golden rule, the second rule of engineering, which is to optimize everything. I think we're going to have started using the term dumb smart for our AI because we're going to be incredibly frustrated. It's like how can you be so smart Aaron's so dumb over here? How can you be so dumb and you're so smart? And there is this idea that because there are different dimensions to them, they're not. They're not all maximized in
one dimension. Main thing
is that we're going to make many hundreds of 1000s of species of AIS, and we're gonna engineer them to do different things. And they're all going to be different and different from us. And the main thing is the poor this AI says that there is not as the ai ai, there is a eyes. So we can train ourselves to understand that you're going to be making cuts with medical qualities, dependencies. And then we're going to have to come up with the kind of ecosystem of all these different kind of connected cutting machine types. So the good way to think about it, populate space of all possible blinds to many different kinds of artificial minds in this space, the kind of analyzing some and then there's our existing human intelligence and we're going to be probably often in a similar we're not the center is kind of where the central and the suddenly fall from our lessons in order with this new gravy failure, kind of thinking and condition that was involved on this planet to survive, certainly considered conditions and is considered a failure, given all the other possible minds to think it's sort of like worth shouting about the center of and so I think a good way to think about these things. Including artificial aliens, these other kinds of mines have been quite
sophisticated and interesting. Maybe we
better think about them as aliens, the related debris, we're going to make ourselves different kinds of things. And so the important thing to say is, is that they are not like huge, don't think like us, they think different. And in a world where we're connected together, come to our today susceptible to groupthink, thinking different disease, the engine of benefits sent in, well presented is the engine of everything. I think it becomes harder to think differently when we're all things together and the A eyes are going to help us to think differently. You can have different mixes of them and everybody will have their own sort of suit their own their own kind of accommodation and helping them to think different because that is what we're trying to do. So the fact that they don't think like this is not a problem seeks a different course of what we're looking for in this kind of economy
is
the point where they are different human said is it an asset or liability and we make income to what you desire, because they are these intelligences that are not concepts because consciousness is in some ways the liability that we had and advertise them as being consciously
because you don't want him to talk to you what is
hard to focus on driving and not to be worried about whether it should really majored in five minutes.
So mostly, I'm going to be telling you something we only
want in certain case. So far, all the work we've done and all the exciting news is really based on the fact that we only been able to synthesize one time they are or what kind of cognition, that type of cognition is basically, pattern recognition and pattern generation. It's only one though different kinds of thinking. We have not yet been able to synthesize, deductive reasoning, intuitive, symbolic reasoning or logical deduction, all these other kinds of modes of thinking we have been able to synthesize and get. So all the things that we've seen in all the things that we're experiencing right now. Progressively, this has been one type of partnerships and that pattern recognition
pattern generation.
And so that pattern generation is the
neural nets so deep learning,
and it has taken us further than we thought it could go, but it is very likely to take us all the way back to wherever. So, but so far, just with that kind of cognition need to generate things like these AI images which are basically
taking shooting with you have had this environment
this cat miss you know, ladies this barbarous and combined with these images. Here are some images of items that are generated images for almost a year, every day making things and this is an example of the kind of artificial creativity that we are capable of.
I say that artificially created in the lower
case creativity meaning that they're tired every day. They're not grateful. This is not the capital of secret keeping busy
really cherish
a lot of our innovations. But that kind of artificial creativity is now the time is something that I have access to somebody who did not believe could have done it. But we have every clear indication that you can generate is lowercase to get to the kind of person that say slavery designer professional level designer would want it to be making a logical every day. It's got to cost us
tremendous later. It's everyday you're
going to usually miss something.
And that's the kind of capability that these
guys have
and what we're seeing is almost instantaneous for Germany Incorporated. So the same kind of pattern, generation kind of recognition is now a plug in for Google
Sheets to the spreadsheet. So your spreadsheet now can
do these patterns as the built right into
Photoshop. So within Photoshop into this,
and Salesforce slacker all as quick as possible just for completeness and visibility building into their tools. So you have so it reminds me the late 80s, early 1990s When there was a huge excitement in the first point, and I literally every day, there will be something new and different. That was installed unlike say crypto there were things that we actually use and change your daily habits. And actually, this is very similar kind of excitement.
But now that we're gonna be seeing
the next decade where literally overnight, something new
coming out, and we can be corporated into our lives. So what's interesting is that I have a sequence of people who invented it and buying a large investment
really knew that they're going to take over there are a kind of specifies
is required, and what's happening is the people using it like us are actually discovering what's good for right this is this is one of the messages which is that during use discovered the benefits and harms of these things, not by thinking about them. It's almost impossible to frequency that so we're the excitement is that there's a millions of people have discovered the one thing that
they can do now
and that's the question of whether before excited
interviewing and training and polling and so people are using these on a daily basis.
What are they using these things for? And one of the things we know that they don't want to do.
So these kind of references can be put into practice so they can go
down on the field and using TTS as well. I can actually identify individual lettuce plants
one by one and recommend or by some European vessel yesterday. I'm going to categorize those into healthy when we water the fertilizer, plant by plant that's called precision agriculture. That's something that a farmer would like him to
pick so that I can transfer agriculture to the pattern initiative.
And this is a job that we don't do.
So I think it's kind of productivity and increasing efficiency of things
is something we've been using this game for a
lot of time is what we're gonna use AI for as well. But competing with a robot
is a terrible metric for humans. Often you know,
we might jog for credit is not that important. And what kind of jobs are those three? Well, things like science, science is inherently inefficient. You have to fail. Japan skip experiments that don't work in order to really learn anything with every experiment and during a
scientist work. You're just not
trying like you're not really learning new things. So science is inherently interesting innovation
is at least this is interesting because you're trying to go work with a
prototype and then you have failed one day. exploration by definition. before it's done, dance
is not as easy as hard, breaking the cost of the payments for hours and it's not about efficiency. So the things that we value as humans alike are inherently inefficient. Putting the small call and things like that. And so we're good at that robots are really good friends. And so we're about whether or not that's going to kind of battery management today guys knew a lot of things really will procure detects I think there's gonna be four models for frameworks. We have nothing about us and one of the contributors slays nurses treat them as pets, animals to have that kind of relationship and then later we need to consider the spiritual gods and the last one is able to set targets intelligence. So the I think that's, that's going to lead from the stance of treatment of slaves. Incredibly growth and that that's not I think that's a possibility but that would be something that you should steer away from. The second one is pets for animals. These are not people to production, the production testimonials to see how stable because feels like your pets and they will there's only a few minutes they will begin to take on some of the attributes of testing and for some people that really good stance relationship. Spiritual gods are sort of a singularity is talking about Terminator within pencil powerful would tell us he does. It's I think that boxing also is not really very likely perhaps the one that I think we're gonna go with these smart fighter aliens. We're gonna have super happy customer consciousness associates, but they're going to be there and it'll be useful to us to work together. Okay, that is an important part is that they become assistants, co pilots, Central US, team, team members coaches, guides, all those kinds of reciprocate desperately we're
in with this. So
that role is what we're seeing already happened with Genesis ar, ar pretend to check that they're working as a joke join us. We're not going to take over us.
We're not going to enslave them. We are partners and
we have a different kind of relationship.
So I wanted to be versatile personal content.
The blank page one
of the things that we're finding out about API's is that they're really good at helping getting
over in my case, the first draft the first step they're going to
make the first you're going to have some bullet points. It will be very very embarrassing to lose their marketers rules
and so
you work with us and we help them in their
in our systems in their house to us. This is by the way, I worked with my image generator and so what did he was using so right now, the listen to marvelous exciting moment where again if we're discovering all abilities to work for a few new infections did not know about and so I had, interviewing asking people who were using these on a daily basis before
dying to work through the building process if I can if you can decipher because I have some items that are really very boring as
these guys into the building those spot building this is a this is the scriptwriter, who gives us scripts to kind of evaluate and give the third eye on whatever steady forces producing the plan complex
problems, errors and loss of especially cases
producing product descriptions, you can leave those on and
every hour is somebody else's discovery as simple as that. Again, they're not they're not producing the final work. We're very involved. In how to produce the best we've ever done regard some experiments where they've shown us a
full set of the programs are
56% more productive when they're releasing these
API's to help them recover inviters complete past 36 7% faster when they have
access to these tools. So another way that we're using that is fill in the blanks. So there's a lot there's almost no interest during this meeting to substitute the crowd out or humanize most of the stuff that is being used whether it is no Art Hall.
I had an assistant committee
for years with my right surgeries down and she chairs and
now she puts in a few when they are illustrations for things,
which is a domino
bites in the past. And so
there's
a place where there was no art and now there is art. And that's often where art is doing is in places where there isn't a call. So it's filling in the blanks. And that's often the true level of tax, and other parts of the tech
DPI which is generating stuff when there wasn't anything at all. So it's not really displacing this. This
is filling in the blank spaces. I did industries and I try
to find out
how many AR games is. There is 30 million distinct brand new images generated every day and most of those exist will never be seen again. That's one of the things about you. It's required to get back to them to get so these images are appearing and then go forever. You've got fishing possibilities, so 99% of them have an audience of one being separated by law for the pleasure of the person making is almost like the same kind of question you might have with walking into a grave in the Netherlands backlighting, you hiked up the mountain you there and see this beautiful thing, and that's why you came you came to see it and maybe no one else to see it and that's fine. And that's a little bit of what these AI theories people are generating myself, just because they're just like
having my own apartment.
And I get to see the pleasure of seeing
what they feel about AI
and how it works. It says like moves with the crowd kind of knowledge and go about dealing with the candidate. So the idea is there's counties Arizona State cap county buildings a big job, and you win the prize and what they discovered was that the most accurate count was the average of all the guesses and that's what we get with the ARS we're getting wisdom of the crowd kind of knowledge is sort of like it's like this the average that's how it is traded on the average walk in and out for the best and the worst. And it's given us a very kind of way mediocre good enough average response date and so that can be very useful. Again, it wins the prize at the county fairs that is not necessarily what we went through our office. And so what we're fighting for people who are using it is that in order to get the intern to do better is that they have to push that off. You have to push them to the be a little bit more tentative but actually what they're discovering is that if you want to go by if you want to say that you are a commodities broker and God you know in some other countries you have to make some decisions in a hurry. And so there's there's kind of role playing that gets them to get out of the ordinary, that is their default. And there is this idea of this kind of a slider between an average and greater. And the problem is that slider is kind of also corresponding to another slider. Which is the truth, house nations, that they have a very good way to do especially when they're pregnant. But oftentimes, we don't want them to be pretty much stick to the facts. And so we're learning how we can steer them to say, the truth and
in this case, we want to know the facts and help them to make stuff
but this is to be creative. Try something that's really strange and different and on the push in
that direction.
So what is this time? This is kind of generated
the artist we're talking about are really good at is synthesizing between two very different places. So we take what's between what about jumping spider versus thin
air
generally, that is very, very difficult for us to do on a regular basis.
We can do that. But we
don't often get into so much trouble.
They go by doing it in seconds again and again and again. So you kind of waste in that sense, right? So it's like, yeah, I should hire ours to do that. But we do a lot of work to kind of take wellness, jumping spiders and coming in and kind of understand both of those environments. And so that would be a big project, but we made it two seconds and that's really marvelous. So you see the same thing in the text, where there are different very low different kinds of things you can ask can be tempting to to actually take the tender out contest and you are going to have to design a way botanicals structures,
what would that look like?
And then they're able to message or study all the technical
patterns last year when the facts are defined and make something that's based on those.
Make some factors in the chemical past you can do that. That's something we could do. But we will not like to do more than one of them. And the intern is able to do a lot of them overcome to this idea of in the between states. Very, very, very good. Example would be 25 examples of what would happen if you don't have an incentive to do that we have that has really acknowledged them extremely well and we know the technical aspects but maybe you can say is that has something to do. And as I said, this waste it's like we wouldn't do that just wouldn't be so much trouble. We're kind of wasting a lot of a lot of the early days of search in search first came along before Google was offerings. And I was trying to do search in the 80s somebody for dialogue because he was he was waiting about 25 hours a day to search for something $25 So new money went into search the long list of everything you need to do when the way to search for combat. He would never waste a search like searching combinations that will just be wasted. You would have to something
frivolous without searching for recipes.
It was very, very expensive. So what search brought from us and societies in these ways to search all day long. Wherever we wanted to. This was a community power and from that we built into the price and then we're going to say the same thing now that the synthesis between things and we're just gonna waste your kind of
very, very hard to do before. Now that we were just going to do this here and we're going to kind of have an abundance of synthesis.
So there are a veteran, they are dispersed nearly.
They spent 1000 hours
whispering to these guys in order to find out to understand how they work, change in order to get the right spell to train other people to get better and some people are really,
really 10 times better than others. I spent hours and hours making this up by measuring and it was not just it was a conversation back and forth. trying different things while they are writing photography for stalking. Very cool. I think this is kind of promising. We see what's right here. Oh, so we'll try this place over here and you are together exploring previously and I have to be very positive about putting my name next to it
as COVID. As I said, the people who do this really well,
are now called confidently years of practice. And they are sometimes 100 times better than the average person because they're spending so much time and recently there's been three jobs. We're constantly hearing already on LinkedIn and whatever is offering salary meaning when a certified people tell about what you're here to do from here.
whisper to you. So
if you're going to pay back earlier,
yesterday, that's the girl for a fantastic
job seems to the legendary therapist, and he was talking about some of the data bought her then she was kind of discussing her feelings about it and the fact that she's the therapist and this was someone who was taking all of our podcasts and put it into these machines and made replica replication occurs at the bottom to interface with the therapy. And the person said that was really helpful for that. And as I said this actually wasn't bad but it was kind of added to the
effect information and the feedback that is left estimates blocks.
Okay, whatever currency whether Hodges says do not add these human, new ones that in a situational awareness, the new black hat had a chance to show
it was different in that very, very successful way to where you're at in some kind of
artificial intimacy, rather than the kind of full blooded, very high dimensional relationship that you want. To have a few minutes. That's all true. But that's the rates of books that are not a price that people read that are not at all aware of the situation, context of the person who is reading. And so in some ways, first of all has to be better than the book because at least it's interactive. And there are people who swear by the therapeutic effect of having pets and dogs. And it's very clear that the dogs from therapeutic responses may not be the optimal thing for for humans, right they may we could be wrong about the battle. Just like your thoughts and practice going to be fallible, so in some ways, but they did miss her body, her body better than a dog. And then there's also the parts of the world where there are people who don't have access to God or human therapists, where this bypassing did better than no benefits at all. So so so I would say
that the best the best advice is actually better than
better than it's better than nothing at all. And that's what we're finding out from a lot of these things. Is we always get have less compared to what compared to a human. Yeah. Compared to nothing, or none. It's a huge advance. So there's a reaction against some of this. The end there is no AI generated images, some parts of our public life or that and places where they're already clearing houses and sharing our lives. We wanted to prohibit people from AI and M. There's also a recent challenge where somebody tries to copyright a book with all the AI are in US dollars, and I decorated it was denied by the copyright office and said that's not copyrightable. It wasn't made by humanity. So here we have some a picture that is claiming to be uncopyrighted. This is obviously can't be sustained. So I think I think that the copyright is long in the bottle for trying to deal with some of these issues. I haven't suggested that there was another kind of like something you might think about as a reference, like whether somebody is being referenced in the making and training of these things, not just whether it's a coffee, because coffee is just or not the right model for this what we want to have some way to measure the influence of piece of art in training the AI is mix of art, but this is very, very, very hard for me to have. Everybody here has been influenced by many people in our lives. And we harm ourselves I suppose, where the influence again and you don't want anyone's house to have a particular thing. And so it's a very, very fraught enterprise to try and measure that. I think we've actually picked them up to have some more ethical training you guys that have been heard during the training set was much more of a selected an opt in idea for the artists to be included in the training set. But in fact, I think there also may be other sets where people are
clamoring to be included in the training set because that's sort of influential.
And so maybe there is some way to to send out some of these tests and, and gains in learning but the influences are generally good and so all that being said about still images I think the real superpower is not in those making photographs are still images. I think we're at peak to the the real revolution, the real question we're going to real superpower to be at least is when we go into change. Can we make 3d models and movies? So you are I here to probably make up our low typewriter size age of drawing something if you've ever had a dish, but none of us like the 3d world or so it's a million dollar kitchen. But could we make a 3d kitchen where we could walk into by ourselves in a day and in the future we're here to write a book and that book will be very similar to these basic people can be themselves right and well, you know, you hear you can make a movie by themselves. But these tools will feature versus of them. could allow us these interns to assist us to make a movie ourselves and one person and they already have examples that means they eyes being used to generate video. Okay, so already these tools are starting to happen. And so and then the metaverse has to an immersive game world, with our way, way beyond a single person, but are actually feasible to happen within the powers of one person. So for some work, you have an idea in making an entire thing in the world and your ability to apply it and that really huge transformation in our culture. So these I decided soldier losing faith in the way that we have a solar generated song and album solar generated novel. You could suddenly have that and that was going to be very, very very powerful. The other way that AI continues very, very important is that AI allows us to search inside a video. Okay, so what I think is that right now if you do a search on YouTube, some of you're searching for maybe the titles, maybe the
captions,
but you're not really is that what it actually was in the video. But let's say I gave the AI an assignment to show me anywhere on YouTube where there's a person who opens the door and was like kind of door that means kind of symmetrically understanding of Parsons all in the bureau. And that can be very, very powerful because it gives us access to this new cultural center. I think the culture of our center has moved from books to moving entities used to be people in the book, now are people in the stream and so being able to have access to and search and manipulate and understand hearts inside of the videos intended to be tremendously powerful. And that's why up to help out Google by now in a race for the search chat TVs because they have access to the inside. And that's a very, very big asset. So technology succeed when they become invisible. And I think that 99% of the API's that we're going to be having a whirlpool will never see it on
the back office is doing
and that means that they have succeeded. So what I've been mostly talking about the front facing but most of AI is going to be prior walls back offices during this thing. And only two were ever coming to me interface. That interface is very important. So most of the time has happened so far. Today is because not to get the air out here. What's new, is that we can't have a conversational interface. We have fears and that they are they're responding to us that are available that listen to the language or for the longest time. It's a glorified model attitude existing pattern. has suddenly become a use case. And you might think that
we're not someone doing life changing. You're typing
into the Google textbook. Everybody ignored it. Nobody was depressed. News. There seems to be no revolution. But then the webpage was visual. That visual interface transformed everything together, because there was an interference revolution. And that's what's happening right now is this is an interface, which is where the AI has been around and now we have this thing of conversing with and we're going to test with keep expanding that so we'll have to manage everything. Everything's gonna have a better conversational interface, so that we can have our hindrance and you have the ability to start off with just take something and add this conversational interface to it, that lack of evidence. So
that's not we were really really shocked from getting that motion through again. It was most something that we have to have
to have but most of is actually going to be very, very primitive. It's actually pretty easy to program and most particularly when we had this conversation. And I don't think I'm prepared for that. The multiple combine to the to the
private key primitive was a very useful tool for setting optical flow person you're talking to us and have so
chose to we're going to kind of ease into that from that time because it's going to be a shock to us. Last thing about the devices is that they are trained on the average. As I said, the best of the worst of all humanity is what they're looking at, and they are as incarcerated as humans. And so are they racist, sexist? Yes, because that's what we get. But we're not gonna accept that. You don't accept the fact that they are, on average like us. We're going to demand that our AIs are better than us. Didn't have you better than we are. They can't be as sexist racist need done whatever it is beyond that. We want them to be better. And it's actually not that difficult to program into the eyes, that ethical guidance and stuff, just just code. And we have some ideas, as most news, robotics or some kind of as a starter solution, we can do thing The problem is, is not AI. What is human awareness is the fact that our own human ethics and morals are very shallow, very inconsistent, very, very dead. And when we try to teach the AIS to be more ethical a little bit we realize that we are not very consistent in this convention to actually try to imagine so we can make them better than us. We don't know what the upside is. That doesn't mean we're into something that is what we mean by us. So so that's the real challenge. And as he tried to do that, to think about that, and trying to create that, that process, I believe this is this is going to cause this way 100 year then ejected places. We're also seeing the same thing happening, it's going to change, who to reply to, who should be given the size, and status. And so this is going to be the intended AI. This
company is not just about us, it's about our own identity.
Who are we
who we want to be one word that is enough.
So I think that the human purpose of AI
and robots is to help us
become better patients. Let me give you the final
takeaway lessons from my view on this. You won't lose your job, everyone now has a big thing with ALS
and pay. You have to crisp up in terms of your great work with them
together. There really been a synthesis and just wait for the habit. So
what would you do? With your own so far? Thank you. I have a book coming out.
