Artificial Intelligence What s next

Show video

The adoption of artificial intelligence  stagnated in 2018 when just about every   second company was using it. But things  took a rapid turn In November 2022. That’s when Open AI released ChatGPT, a chatbot  able to generate human-like responses. It became   the fastest-growing consumer app in the  history of the internet. Within two months,   it surpassed 100 million users and its  servers were frequently at capacity.  

Search interest for terms like ChatGPT,  AI, and Generative AI have skyrocketed. People in the tech world are now running  around like chickens with their head cut off,   which is as good an explanation for  why the chicken crossed the road as   there’ll ever be. And this is only  the beginning. In this video I want   to look at what’s next. What are startups  working on, how will it change our lives,   and what jobs are likely to suffer. Where is  AI going? That’s what we’ll talk about today.

First things first, Artificial Intelligence is  a catch-all phrase for computer systems that   can perform tasks commonly associated  with human cognitive functions such   as interpreting speech, playing games,  and identifying patterns. AIs are often,   but not always, modelled on ways that  the human brain learns or evolves. One of the things that human brains are  reasonably good at is understanding and   replying to written and spoken language. This  Natural Language Processing has for long been   a stumbling block for AI. ChatGPT has demonstrated  clearly that this obstacle has now been overcome. Unfortunately, we know little about how it  works. The company OpenAI was founded in   2015 as a non-profit research lab by a group of  investors including Elon Musk and Peter Thiel.  

In 2019, Microsoft invested one billion dollars.  Following ChatGPT’s stunning success, Microsoft   wasted no time strengthening that partnership,  reportedly investing an additional 10 billion   dollars in January this year. They also swiftly  integrated ChatGPT into their search engine Bing. One of Bing’s first missions was to try and  convince a New York Times columnist to leave his   wife. It didn’t work, and Bing has since learned  to not ask questions. Google quickly got into the   game, too, by presenting its own AI-assisted  search engine called “Bard”. Unfortunately,   a demonstration video shared in early  February contained a blunder about the   new James Webb telescope. Google’s stock value  promptly tumbled, though it’s since recovered. OpenAI originally planned to share patents  and research insights, but it seems that once   they realized just how much money there is to  make, they’ve reversed course. Ilya Sutskever,  

co-founder and lead scientist at OpenAI, recently  commented on this lack of disclosure, saying that   the landscape has proved too competitive to  reveal specifics on ChatGPT’s architecture,   training models, and dataset construction.  Funny how money can change your outlook eh. Not only do we not know how it works, we also  have no idea what this sudden development   is going to do to society. The situation has  many people both inside and outside the field,   including Elon Musk and Steve Wozniak, so  worried they’ve asked for a pause with further   AI developments. In an open letter that appeared  late March they wrote that “recent months have  

seen AI labs locked in an out-of-control race  to develop and deploy ever more powerful digital   minds that no one – not even their creators –  can understand, predict, or reliably control”. Meanwhile, people all over the  world are trying to find ways to   put chatbots to use. I’ve discovered  that ChatGPT is unfortunately pretty   miserable at writing YouTube scripts, so  for the time being you’re stuck with me.   But some obvious uses for chatbots that  aren’t hard to guess will become very common   are producing social media content and writing  emails. For these and many other applications   it’d be preferable if the AI was trained to  emulate you personally, not just any human.

That’s what I think will become  the dominant application of AI in   the near future. ‘Personalized  AI services. Machine learning   algorithms that analyse and learn  from your feedback, your behaviour,   your speech, your preferences, and  your habits. Software that groks you. We have already seen the beginning of  this with recommendation algorithms   that suggest anything from the next  video to the next romantic partner,   what’s the difference anyway. But now make that  every-day decisions. How do I fill in this form?   What should I have for lunch? What’s this thing  on my you-know-what and do I need to see a doctor   about it? Everything you ever wanted to ask but  didn’t dare to, answered by the most patient and   understanding companion ever. Your personal AI,  like personal Jesus, but one who actually replies.

One new startup that wants to help you with  this is in fact called Personal AI. They’re   close to launching the first version. The  app’s a messenger that you train on your   knowledge about the people in your life and  that’ll then help you interact with them,   or, in fact, do the interaction for you. It’ll let you create profiles for connecting  with different groups of people: work, friends,   relatives and so on, and help you communicate  with them. It can even answer on your behalf,  

and it won’t be long until  our personal AIs are having   the most interesting conversations about  us but without us. The future is bright. You may think this sounds like a software  manifestation of multiple personality disorder,   but for people like me who are really  bad at hitting the right tone in social   interactions, it’s going to be a  blessing. Personal Jesus, indeed. If you don’t just want an AI to talk  instead of you, but to talk to you,   then maybe you should check out a personalized  chatbot. These have been around for some years but   they’ll without doubt see major upgrades soon. Let  me just pick one because it’s an interesting case.

The Replika app was first released in March 2017.  Replika chatbots have avatars and learn from the   user’s input. They provide emotional support,  companionship, and entertainment. Users can   update their mood within the app, and the chatbot  will adjust its responses based on that. If a   user is feeling sad, for example, the chatbot may  offer words of encouragement or activities to help   them feel better. If they’re feeling happy, the  chatbot may respond with jokes or playful banter. Replika is also a cautious example. It used  to have a subscription-only option for erotic  

roleplay. In February this year the company  received a warning from Italian authorities,   among other things because they didn’t do enough  to make sure underage users were protected   from improper content. Without warning, Replika  removed their adult features basically overnight,   leaving many users seriously distressed,  reporting they felt like they lost a friend. What’s going on? Here's how I think about  this development. Our options to change  

our own thoughts from within are limited.  This is why we’ve for long used externalized   feedback to improve our mental health, such  as writing a journal, talking to ourselves,   or actually seeing a therapist. It helps because  it’s a different input than internal speech. AI is yet another method to do this, but it’s  a method over which we have limited control.  

If you accept software as a friend, even though  you know it’s not a person, because that really   makes your life better, then the pain when they  leave will be equally real. I don’t think that   anyone at the moment understands the psychological  problems that can be created by personalized AIs. In the future such personalized life-managing  apps are likely to have integrations with other,   specialized apps, for example for medical  or legal advice. Several of those already   exist. For example DoNotPay is the  first artificially intelligent lawyer,   and ADA and Babylon Health give medical  advice. And this is all well and fine,   but no one wants a different app  for every niche of their life.

Another improvement for your personal life  may be quickly finding that document you   remember reading last week, but where is it  now? There’s an app for this. It’s called   Rewind and it records and catalogues virtually  everything you do on your computer. You can ask   it about that thing with the guy who brought  the stuff and I’ll use its best artificial   intelligence to figure out what you mean. This  app’s been around since 2020 and currently   only works on Mac computers, but you can bet  we’ll see more of this for other systems soon. Those are some of the changes coming to  your personal life, now let’s look at   art and entertainment, where the impact is  huge already. AI-generated art isn’t new,  

but it’s risen to an entirely new level with  DALL-E and midjourney. They can convincingly   create artworks which are at first and second  look basically indistinguishable from real art. Already in September last year, a  midjourney-created image won first   prize at the Colorado State Fair’s annual art  competition. A similar thing happened a few weeks   ago but this time in a photo competition. And after a recent update, midjourney   seems to have learned that human  hands usually have five fingers,   so there’s nothing stopping it  now from taking over the world. A lot of artists aren’t happy. Would you  believe it. Artificial Intelligence gives  

everyone the ability to create art from their  intention without the need to have learned the   techniques. That’s great if you haven’t learned  the techniques, not so great if you did. We’ll   without doubt see a lot more of AI-generated  art, but I also think there’ll be limits to it. The next area that’s likely to blow up is AI  animation. Production studio Corridor Digital   recently unveiled a short anime called Rock,  Paper, Scissors that used AI to learn natural   motion and three d panning from motion pictures.  The production studio was criticized by animators   and other artists who complained about the lack of  artistic value and originality. You’ll understand   if you watch the thing, but I think such criticism  is missing the point. This short animation is a  

first warning for how AI will alter the film  and animation industries in the years to come. And then there’s streaming. Faceswaps are  yesterday, today we have AI-generated streams   and television shows. The popular streaming  service Twitch is now host to several AI streams,  

like ai_sponge247, which streams AI-generated  Spongebob episodes 24/7 or “Nothing forever”   that’s an AI-generated parody of  the American TV series “Seinfeld”.   Twitch also already has AI bots that mimic  popular streamers. Users can ask questions,   and the AI streamer will respond using the same  style and intonation that the streamer uses. It   still looks and sounds a bit wonky, but you can  bet it’s going to improve rapidly. I frankly  

don’t understand why people watch these things,  but then I also don’t understand why they watch   my videos. And in the end it doesn’t really  matter I guess, so long as they like doing it. Eventually we’re going to see full AI generated  videos from text prompts, like midjourney   generates images, so that’ll be quite a trip.  And then there’s music. There are several AI   based software solutions that create new music  from text prompts, for example Amper Music or   Soundraw. These typically let you enter a mood,  genre, type of music, and so on, and will generate  

a royalty free soundtrack that you can use for  videos or podcasts. This “music” isn’t going to   win any awards, but it’s good enough to run in  the background and there’s a market for that. There are also already some entirely artificial  musicians. Here are for example Yona and Miquela. In the future, artificially enhanced music  production is certainly going to become   more ambitious. It’s not much of a secret  that popular song writing follows simple   and predictable patterns, so AI is bound  to have a big impact there. It’ll also be   really handy for writing lyrics, especially if  those don’t have to make a lot of sense which,   let’s be honest, is the case for most pop songs  anyway. Yes, Lada Gaga, I’m looking at you.

We now also have AIs that emulate singing  voices, and do that really well. This is   why you can now listen to Kayne West singing  everything from Coldplay to Justin Bieber. In the past two months or so, style mashups have  begun to appeared on popular streaming platforms,   leading to a wave of copyright complaints. Google  has developed a platform for AI generated music.  

They have written a paper about it and have  examples online but they haven’t made the   tool publicly available, probably exactly  because the copyright issues haven’t been   resolved. It’s basically like midjourney  but for music. Here are some examples. But having spent some time on music production  I think there’ll be limits to AI use in the   business. Some instruments and audio mixes are so  complex that it’s difficult to even explain what   you want to do with them. It’s one thing to take  an already existing top song and tweak the voice,   it’s another thing entirely to create  it from scratch. This is why for the   most part electric guitars you hear in  pop music are actually electric guitars   and not computer-generated audios. This  is also why many synthesizers are still   hardware-based. It’s not because  the hardware is necessarily better,  

but because it’s faster and simpler  and easier to deal with than software. So this is where I think the limits of AI use  will be. If it’s more difficult to explain what   you want than just doing it yourself. But  voice generators do have other uses. We’ve   seen this on YouTube for a long time already  that people use AI voices to dub videos. You  

can now also train AIs on your own voice  and then use it to create further audio. This for example is gibber I didn't read. I gibber  a software called Overdub and then just entered   the text. It replaces every other word with gibber  because they want you to get a subscription. I suspect we’re soon going to see a lot  of this for automatic translations in the   future. Chances are in a few years from now  you’ll be able to watch this video in German,  

with an AI generated voice and translation. So  if you make a living by reading audio books,   I think you’ll soon have to look for a  new job. AI generated voices also open   entirely new possibilities for spammers because  they can now call with your grandma’s voice. Okay, let’s then have a quick look at work  life and the business sector. The biggest  

impact in the business sector is going to  be in web design and software development,   and it’s happening already. That’s because it’s  a combination of language processing and visuals,   and AIs have gotten incredibly good at both. By using ChatGPT’s newest version 4.0 you can  basically create web pages by giving speech   commands. You no longer need to know how to code.  Yes, that’s right: You tell an AI what you want   the website to do and to look like, and it'll  write the code for you. Just look at this guy. Yes? I need another next app with Tailwind One sec.

You want me to create a new  next js app with tailwind   css? Yes One moment What would you like the app to do? So this time I want a basic social   networking app and it needs to have three  things. It needs to have a profile creation form,   it needs to have a profile viewer, and I also  want a way to see all the users on the network. One sec, I'll add those fields to the  profile schema. What else can I do?

I want you to optimize the site so that  it works with mobile and desktop devices   and I also want you to style  it in like a dark mode. Okay just now it's building it's building. Boom.  Dark mode. Let's see if it's responsive. Okay,   well it looks fine. The game has  changed everyone. This is wild. If you stick around for a bit on the midjourney  servers you’ll also see that people frequently   use it to “imagine” webpage designs or logos  for one or the other purpose. It isn’t hard to   extrapolate that soon a startup will combine one  with the other for personalized website design.

Of course this isn’t going to make software  developers entirely unnecessary, because AI   generated code will sometimes not work, and then  you’ll need someone to sort out the problem.   However, I think we’ll see a shift much like the  one we saw 20 years ago from writing websites in   HTML to content management systems that create  a website with one click from a template. It’ll   be imperfect, and sometimes annoying, but for  many purposes it’ll be good enough. And it’ll   mostly be a good thing because there are  a lot of really crappy websites out there. Another application of AI that has many uses  in business is the automatic identification   of objects from images and video. For example,  the startup Voxel offers software to monitor  

manufacturing and industrial facilities  with the purpose of identifying safety   risks in real time. It’s already being used  by companies like Office Depot and Michaels,   and has in some cases reportedly reduced  workplace injuries by 80 percent. Another example is a platform called Viso Suite  which offers the newest AI-driven object detection   models to incorporate into your business. It  can for example be used in retail to find and  

track products on shelves, or it can be used  in manufacturing to detect defects in products. AI supported image detection and analysis  is also being used in many health care   applications already. For example the company  NVIDIA has created a service for the healthcare   industry known as Clara. It can be used,  among other things, on multi-organ scans  

to separate the data into single organs and  then create comprehensive visualizations. Of course, such software can also  be used for face recognition,   which brings up a lot of privacy  concerns. Do you really want a face   recognition software to track who walks  in and out of your hotel room? Right.

AI is also going to have an impact on academia,  for example by making it easier to find papers.   Elicit is one of the first to try it. It’s  a free app from a non-profit by name Ought,   and it uses natural language processing  on a database of 175 million research   papers. You can ask it a question and  it’ll bring up references. It’s still   an early-stage product but new updates and  improvements are being rolled out weekly.

The potential of AI driven analysis of  the scientific literature is enormous,   because it’s almost certainly the case that  some questions have remained unanswered just   because someone couldn’t find the paper  in which their problem had been solved   already. AI can do it because it’s  ultimately just pattern recognition.   Once AI is able to identify abstract  ideas expressed in graphs or equations,   a lot of connections are going to be made,  which could lead to a lot of sudden progress. Many people are concerned about the  sudden rise of AIs, and it’s not just   fearmongering. No one knows just how close  we are to human-like artificial intelligence.   As I’ve said previously, I have no doubt  it’s possible that computers will one day   be conscious and quite possibly more intelligent  than we are. The human brain excels in efficiency,   not in function, which makes it plausible, indeed  probable, that if you disregard efficiency,   the functionality of the human brain can  be much improved on. This could solve a   lot of our problems very quickly. It could  also *create a lot of problems very quickly.

Current concerns have focused on privacy  and biases and that’s fair enough. But   what I’m more worried about is the impact  on society, mental well-being, politics,   and economics. It’s extremely foreseeable  that the forest of new AI startups is going   to thin out rapidly and they’ll end up being  subsumed in a few all-purpose apps that’ll   dominate the market. And when hundreds  of millions of people are going to leave   every-day decisions up to a few AIs, even  a small mistake can have huge consequences.

But that’s probably not what most people  are worried about. Chances are they’re more   worried they’ll lose their job. And that’s  indeed a reasonable concern. A just-released   report from Goldman Sachs says that the  currently existing AI systems can replace   300 million jobs worldwide, and about one  in four work tasks in the US and Europe. According to Goldman Sachs, the biggest impacts  will be felt in developed economies. Artificial   intelligence will first replace jobs involving  repetitive tasks, from data entry clerks and   customer service representatives to factory  workers and telemarketers. They expect almost half   of all Office and Administrative Support and Legal  roles can be replaced by AIs, while trades jobs,   as well as maintenance, repair, and construction  workers are mostly safe. Until the robots come.

What do you think about these developments? Are   you more worried or more excited?  Let me know in the comments. Yes, artificial intelligence is really everywhere  these days. But just exactly how does it work?   A great place to learn more  about machine learning and   neural networks is Brilliant dot org  who have been sponsoring this video. Brilliant offers courses on a large variety of  topics in science and mathematics. To learn more  

about artificial intelligence, have a look at  their course on Computer Science Fundamentals,   their Introduction to Neural Networks, and their  course on Artificial Neural Networks. It’s a fresh   and new approach to learning with interactive  visualizations and follow-up questions. I’ve found   it to be a highly effective way to understand  and also to remember material. Brilliant lets  

you pursue your own interests at your own pace.  And they’re adding new content each month. I like looking things up on Brilliant,  and I now even have my own course there.   It’s an introduction to quantum mechanics  that covers topics such as interference,   superpositions and entanglement, the uncertainty  principle, and Bell’s theorem. It’s a beginner’s   course that you can take without prior knowledge.  And afterwards, maybe you want to continue   learning more about quantum computing or special  relativity or wherever your interest takes you.

If you want to try Brilliant out, use our  link Brilliant dot org slash Sabine and sign   up for free. You'll get access to everything  Brilliant has to offer for 30 days, and the   first 200 subscribers using this link will get  20 percent off the annual premium subscription. Thanks for watching, see you next week.

2023-05-09

Show video