Episode Categories:
Resources
Join the discussion: Ask Jordan questions on AGI
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Connect with Jordan Wilson: LinkedIn Profile
Try Our Free AI Prompting Course: Register for our free Prime, Prompt, and Polish AI Course!
Defining Artificial Intelligence: From AI to AGI
As the world of artificial intelligence (AI) evolves, understanding its different forms becomes crucial. Traditional AI, which performs tasks typically requiring human intelligence such as recognizing images and language translation, has carved its niche in various fields like banking. A newer, more democratized form of AI, known as generative AI, allows even non-specialists to leverage AI technologies such as language models.
Arguably the true holy grail of artificial intelligence is Artificial General Intelligence (AGI). Essentially, AGI refers to AI that can perform a range of tasks with a human-like intellectual performance - a benchmark for AI that can perform any task without needing extensive training. Venture further into the future, and you encounter the concept of Artificial Superintelligence (ASI), an AI that surpasses human intelligence altogether.
Evolving Predictions: The Journey to AGI
The journey towards achieving AGI has seen its timeline compress significantly over recent years. Once predicted to take at least 80 years, the emergence of advanced models like GPT-3 and Google's Lambda has reduced the expected timeframe to as little as 8 years. Such shifts highlight the fast-paced evolution in the field – an indicator that businesses should keep a close eye on AI progress and its impacts on various sectors.
New Age Partnerships: AGI and Big Tech
Artificial intelligence has caught the attention of large corporations like NVIDIA, Microsoft, Meta (Facebook), and more, many of whom collaborate on AGI research and tech. Such shifts in the traditionally solitary nature of AI development underscore the importance of cooperation in the race towards AGI. The potential implications of AGI on the economy are vast, warranting the attention of giants like Amazon and Microsoft who are investing significantly in AGI development.
Artificial Intelligence Startups: The New Players
OpenAI and Anthropic, two leading AI startups, are making direct or indirect strides towards AGI. OpenAI's partnership with Microsoft and their recent partnership with USAID for federal use of ChatGPT Enterprise, demonstrates the high-stakes partnerships that are shaping the landscape of AI development.
Implications for AGI: Anticipating the Future
With the cost of technology needed for AGI decreasing significantly, and a rise in resources being allocated to AI development, AGI's achievement seems likely within a much shorter timeframe than originally anticipated, potentially within the next few years. This significant milestone could fundamentally reshape the world of work, business operations, and career trajectories.
Conclusion: Preparing for the AGI Era
In the rapidly changing world of AI and AGI, businesses must remain informed, agile, and ready to adapt. As we anticipate the arrival of AGI, it's crucial to analyze how it could potentially impact strategies, goals, and the larger business culture. For businesses – from startups to established corporations – understanding the pace of AGI development could well be the difference between leading the change or catching up to it.
Topics Covered in This Episode
1. Definition of AI, AGI and ASI
2. Evolution of AGI
3. Impacts of Advancements in AGI
4. Future of AGI
Podcast Transcript
Jordan Wilson [00:00:17]:
When will we achieve AGI, artificial general intelligence? It's a question that I think about personally a lot, but I think it's something that we should be talking more about because I think the definition of both artificial intelligence and AGI is changing by the day. And I think that there's maybe one kind of secret thing holding us back from achieving AGI. Alright. I'm gonna be talking about that today and more on everyday AI. What's going on y'all? My name is Jordan Wilson, and I'm the host, and everyday AI is for you. It's a daily livestream podcast and free daily newsletter, helping us all understand AI and, who knows, maybe AGI so we can grow our companies and grow our careers. So if that sounds like you, maybe you are brand new here. Thank you for joining us.
Jordan Wilson [00:01:15]:
Make sure to check out the podcast show notes, for more, related episodes and to get to our website. You need to get there if you haven't already. Go to your everydayai.com. Sign up for the free daily newsletter. So, yeah, this is a podcast and livestream, but we have a newsletter every single day recapping both the show and literally everything else you need, to stay ahead, of AI, to grow your company and to grow your career. It is a free cheat sheet, so, you should be going there. Alright. So I'm excited today to talk about when we will achieve AGI or maybe if we will.
Jordan Wilson [00:01:55]:
Alright. Because I think some of what we're gonna be talking about today is gonna be surprising you. So, before we get into it, though, let's start as we do every day by going over the AI news. Alright. So authors are now suing Anthropic over alleged copyright infringement in AI training. So a group of authors has filed a lawsuit against the AI startup, Anthropic, claiming that it has engaged in, quote, unquote, large scale theft by training its chatbot, Claude, on pirated copies of copyrighted books. This marks a significant moment in the ongoing legal battle surrounding artificial intelligence and copyright issues. So the lawsuit was filed in federal court in San Francisco by authors Andrea Bartz, Charles Graber, and Kirk Wallace Johnson who seek to represent a class of authors in similar situations.
Jordan Wilson [00:02:46]:
This is the 1st lawsuit from writers specifically targeting anthropic despite many other lawsuits in similar cases against its competitor, OpenAI, the creator of ChatGPT. The authors accuse Anthropic of violating copyright laws by using a dataset known as the pile, which allegedly contains a significant number of copy of of pirated works. The lawsuit argues that Anthropic's practices contradict its claim of being a responsible developer of AI technology, stating that it that the company profits from strip mining human creativity. Alright. Our next piece of AI news. Google and its Gemini team announced access to 1,500,000,000 free tokens a day. Yes. So the Google Gemini team announced just kind of on the down low on Twitter, and it's making headlines by offering an unprecedented amount of free tokens, 1,500,000,000 tokens for free each day for developers, a move that stands to transform the landscape of AI development for both aspiring and established developers.
Jordan Wilson [00:03:51]:
So this follows closely after, essentially, OpenAI said a similar thing. Hey. It's free until, the middle of September. So, the Gemini 1.5 Flash free tier grants developers 15 requests per minute, 1,000,000 tokens per minute, and 1500 requests per days, per day. That means that the developers can experiment and build applications without immediate financial pressure fostering innovation in the AI space. In addition, this new free tier includes context caching for up to 1,000,000 tokens of storage per hour and free fine tuning capabilities, which are crucial for developers looking to optimize their mobile, their models for specific tasks. Y'all, I feel bad for companies that were super early to this generative AI, boom like 2 years ago. A company doing this would have spent probably 1,000,000 of dollars and not done a very good job.
Jordan Wilson [00:04:43]:
And now it's too I I mean, companies are making intelligence and compute too cheap to meter. Alright. Last but not least, OpenAI has announced or it's been reported that OpenAI has its first federal partnership with USAID to inter, to implement chat gpt enterprise for federal use. So OpenAI has made headlines by partnering with the US Agency For International Development, marking USAID as its first federal agency customer for chat gbt enterprise. So OpenAI's chat gbt enterprise is designed for larger organizations providing advanced analytics and customization features that can enhance efficiency in government operations. So the company is pursuing FedRAMP moderate accreditation, which would enable Chat gbt enterprise to handle moderately sensitive federal data, expanding its potential use within the government agencies. The US White House Biden administration's recent executive order encourages federal agencies to adopt generative AI technologies while addressing concerns about data security and bias. So, yeah, pretty, pretty big news there.
Jordan Wilson [00:05:52]:
Federal government, kind of striking an official partnership with OpenAI, in a federal agency will be using chat gbt, the enterprise edition. And, hey, by the way, if you have chat gbt enterprise, if your company does, do you know we train companies on that? It's not like one of the big things we do, FYI. So, yeah, reach out. Reach out if you or your team needs help, with, Chattopty. Alright. So let's get into it y'all. Let's talk about the big thing here. When will we achieve AGI? And I think one kind of secret relationship or one kind of fine print that might be keeping development back.
Jordan Wilson [00:06:34]:
Alright. So I'm super super excited for today's conversation. So I'd love to hear from our livestream audience. Yeah. Hey, podcast audience, you know, you might get tired of, I don't know, hearing questions from the livestream audience. You might be like, oh, I wish I could get my questions answered. Well, join us. We do this live every single day at 7:30 AM Central Standard Time.
Jordan Wilson [00:06:55]:
So, yeah, let me know, everyone, what are your questions on AGI? Do you think we're gonna get there? What do you think is holding us back right now? Alright. So let's just go ahead and start at the end. I'm not gonna make I'm not gonna make you wait. Alright? But I think right now, we are must much, much, much closer to artificial general intelligence than most people think. Alright. Don't worry. I'm gonna get to what it means, the definitions and the differences, between. Don't worry.
Jordan Wilson [00:07:29]:
But one of the reasons why I think we haven't quote, unquote officially achieved AGI or artificial general intelligence is because the goalposts are constantly moving. Alright? I'm gonna have a little bit you you know, we always bring receipts, y'all. It's it's hot take Tuesday as well. I should have called that out y'all. So let me know. Should I be should I be middle of the road or should I really bring the heat? But here's what I think is holding us back. It's actually some of the fine print and some of the details between OpenAI's partnership with Microsoft. Alright?
Jordan Wilson [00:08:09]:
So that's high level y'all. And we're gonna dive into it now, but I
Jordan Wilson [00:08:12]:
think if we were using old standards, if I'm being honest, I think we would already technically have achieved AGI, but the goalposts are always moving. I think as large language models are getting more advanced, I think the goalposts are moving on what AGI even means, artificial general intelligence. Alright. It looks like Michael's Michael said 3 flame emojis, so
Jordan Wilson [00:08:40]:
we'll see. Maybe maybe we'll keep
Jordan Wilson [00:08:42]:
it tame. It looks like everyone wants a tame show today. That's fine. We can do that. Alright. So let's go ahead and put some definitions out there. Alright? Because, yeah, maybe if you are new to this whole AI scene, generative AI, maybe you're not sure. Maybe you don't know what, AGI even is.
Jordan Wilson [00:09:01]:
So let's go ahead and define it. Okay? Yes. Because the definitions are constantly changing. So let's take a look at artificial intelligence, artificial general intelligence, and then artificial superintelligence. So let's start with AI, and let
Jordan Wilson [00:09:20]:
me just start by saying this. AI is
Jordan Wilson [00:09:24]:
not new. Right? Artificial intelligence has been used in many different industries for decades. It actually goes back to the forties fifties, and it's been widely used by many industries since the seventies eighties. Artificial intelligence is not new. We've had machine learning. We've had, kind of this deep learning phase as as well. So artificial intelligence by itself is not new, but let's go ahead and define it. So these are my definitions too.
Jordan Wilson [00:09:54]:
Alright. So, so artificial intelligence is when machines or software are designed to perform tasks that typically require human intelligence, such as recognizing images, translating languages, or making decisions. And I will say traditional, quote, unquote, traditional artificial intelligence is really based on a set of rules, a set of algorithms, decision trees. Right? It's programmed. There's bits and bytes almost. Right? Traditional artificial intelligence. So like I said, an easy example that's been around for decades, right, is when banks are giving out loans, and they have essentially algorithms. Right? You go into an office, you probably give someone information, you fill out a form, they enter that form, and then there's a an artificial intelligence algorithm that says, okay, Is this person really qualified for this loan or not? Right? Maybe if it's a big loan, you might have filled out a ton of paperwork.
Jordan Wilson [00:10:51]:
So there's a ton of different pieces of data that you're essentially giving the bank. And then the bank uses artificial intelligence to essentially assign different values and to see if you are risky too risky for the loan. Right? So AI has been around for a very long time. It's not new. Right? And you've probably been exposed to artificial intelligence even in your daily lives well before chat g p t. Alright. But, obviously, I will say this. And I was I was at the NVIDIA GTC conference a couple months ago, and its CEO, NVIDIA CEO Jensen Huang, kind of said that, you know, generative AI in large language models marks this this new era of artificial intelligence.
Jordan Wilson [00:11:45]:
And I absolutely agree, kind of this generative AI. Right? So generative AI is a little different than artificial intelligence, kind of, you know, quote, unquote old school. But, you know, generative AI, not to be confused with artificial general intelligence, is different. Generative AI, it it brings it democratizes, right, AI for everyone. Because before, let's say, 2020. Right? Yes. Chat gpt was, you know, released to the masses in November 2022, and I do think that's the turning point. But, you know, this, generative AI technology was available through other providers.
Jordan Wilson [00:12:24]:
You know, OpenAI made their techno, techno, technology available, to third party, developers back in 2020. So pre 2020, you know, you could I say that's traditional AI. Right? Yeah. I'm slapping my own labels on this. Right? Now we have generative AI. It it it is kind of this next advancement of artificial intelligence that lowers the learning curve to, like, 0, y'all. I mean, to take advantage of artificial intelligence pre 2020, I mean, if I'm being honest, you either had to be a specialist working at one of these companies that had niche use cases for artificial intelligence, or you had to be a deep learning machine learning expert. Right? You had to have a degree in artificial intelligence to essentially take advantage.
Jordan Wilson [00:13:10]:
So generative AI, so let's say, you know, the 2020 to, you know, now range, that brings AI to all of us, right, through large language models. And what generative AI is, simply put, is when anyone can simply speak or type to an AI system and get a pretty impressive output. Right? These large language models, these generative AI systems, you know, like ChatGPT, Google Gemini, Anthropic Claude, you know, and then you look at more, image or creative based tools, like Runway, like Midjourney, like DALL E, Adobe Firefly, etcetera. Right? Where you can put in a simple text prompt or speak in many cases and and get something visual. Right? You can get a photo. You can get a video. You can get an audio track. Right? You can get a a voice over.
Jordan Wilson [00:14:04]:
So this new generative AI phase, not to be confused with AGI, has really democratized how the US works and what we can all accomplish with artificial intelligence. But it is but a small footnote in the larger umbrella of artificial intelligence. Alright. So now let's look at what artificial general intelligence is. Alright. So this is a type of AI that can understand, learn, and apply knowledge across a wide range of tasks, similar to how a human would effectively performing any intellectual task. K? That's the difference. For the most part, AI or generative AI, if you will, performs a more narrow.
Jordan Wilson [00:14:50]:
Right? Not even gonna get in a narrow intelligence, but AI, generative AI performs a narrow base of tasks or a narrow set of tasks. Right?
Jordan Wilson [00:15:01]:
Hey. Recap this PDF. Hey. You you
Jordan Wilson [00:15:05]:
know, write this blog post. Hey. Create this, create this image. Right? You're kind of working on one task at a time, and it's a task that, kind of the AI system has clearly been trained on. So AGI is slightly different. That is when a a machine or a system or, you know, who knows, a software. Right? What shape AGI will eventually take. But that's when it can perform a broad range of tasks at the same or higher level than a human.
Jordan Wilson [00:15:37]:
Right? Now let's talk about artificial superintelligence. Right? We gotta get the whole acronym soup here, y'all. So artificial superintelligence or ASI is a more of a hypothetical AI that surpasses human intelligence in all aspects, including creativity, problem solving, and emotional intelligence. ASI would be capable of self improvement. That's a big thing. And could potentially outperform humans in any cognitive task leading to profound implications and potential risks. Alright. I simplified it here.
Jordan Wilson [00:16:12]:
Think of AI like this. AI is like, oh, look at them. Look at what the machines can do. AGI is, oh, the machines are way better than my job than me. And ASI is, oh, be fearful of the machines. Alright? Yeah. That's a very oversimplified way, but that's the way that I think about it in my mind. Right? And, yeah, we've been at this.
Jordan Wilson [00:16:35]:
I think we've been teetering. Right? If I'm being honest, we've been teetering on this AI AGI line, I'd say, for the past 6 months. Right? Between, like, oh, look what the machines can do, and, oh, is this AI actually better than my job than me? Right? And I think that, again, if you are looking at narrow applications,
Jordan Wilson [00:16:57]:
I don't think there's any denying that with certain skill sets, AI is way better, way, way better than a single human. Right? So let's just say if you are a single human and all you do is data analysis, right, all day, AI is way better than the best human out there. Right? But that's a single task. That is a narrow focus.
Jordan Wilson [00:17:26]:
That doesn't mean that you can talk with an AI system, and it can literally do anything and everything that a human can do. That's kind of when we talk about AGI. That's kind of the benchmark. Right? It can perform any task without the need to train it, without the need to necessarily show it examples. Right? We're essentially with a a zero shot prompt or with little training, with little upfront information. You could type or chat or interact with a system that automatically is going to outperform almost the best human on any task. I think we're getting close.
Jordan Wilson [00:18:07]:
I think
Jordan Wilson [00:18:08]:
we're getting a lot closer. So now now let's talk about some of the reasons why I think development and the distance between where we are now in whatever that AGI finish line is diminishing quickly, if I'm being honest. And I think one of the biggest reasons, actually, is now you have the tech titans of the world just racing toward AGI, literally. And let me just call this out. The concept of a company pursuing or contributing to artificial general intelligence, AGI, was straight up taboo,
Jordan Wilson [00:20:04]:
right, 10 years ago.
Jordan Wilson [00:20:07]:
You know, OpenAI was actually you know, in 2015, it was one of the first companies that was saying, like, hey. We're working toward AGI. I mean, at the time, they were a little known startup. OpenAI was not what it is now back in 2015 to 2019. Very few people, even very few people who worked in the tech space knew what OpenAI was unless you had a a, niche in in or around AI or were particularly interested. You didn't know OpenAI or that this small startup, you you know, was working toward artificial general intelligence. But now it's different.
Jordan Wilson [00:20:50]:
It's different, y'all. And let me
Jordan Wilson [00:20:52]:
just say this. Up until 2 ish years ago, you did not have the largest companies in the world openly working toward AGI. Like I said, 10 years ago, it would have been taboo for a big tech company to say this. It would have been controversial. But I think through this, we'll just call it this experiment over the last 18 months, maybe since Chat gpt, was released. Now companies and the US economy sees and understands the value of large language models and the value of businesses using AI top to bottom. So not that it's become trendy to work toward or talk about AGI, but I think that the, the business world has opened its eyes, I think, because they see the dollars. Right? The more you talk about AI in your, earnings calls, the the the higher your stock price goes.
Jordan Wilson [00:21:55]:
Right? But now you have all, not all, but almost all of
Jordan Wilson [00:22:01]:
the, you know, 4 or 5 of the top 7 largest companies in the US openly either chasing AGI or openly collaborating toward achieving AGI. Like I said, this is straight up taboo 10 years ago. Right? If if a c suite executive said, yeah, we're working toward AGI, would have been a red flag. Right? They would have the the the board would have
Jordan Wilson [00:22:34]:
put out a memo. It would have been bad. Now it's almost the opposite. Now if you're a, you know, a tech conglomerate and you're not openly working toward AGI, the board might just say, why? Why not? This is something you should be doing. Whereas before it was, oh, no. We shouldn't be doing that. When we achieve AGI, what happens to our jobs?
Jordan Wilson [00:22:54]:
Right? Let's look at the proof. NVIDIA. They've been the largest company in
Jordan Wilson [00:23:00]:
the world. Now they're top 3. They provide the hardware and software and tools for AI development. They support AGI research through partnerships and infrastructure and their CEO. I was literally in the room a couple of feet from Jensen Huang when
Jordan Wilson [00:23:14]:
he said this. He did say that we would achieve AGI within 5 years. K.
Jordan Wilson [00:23:20]:
Let's keep going. Microsoft. Microsoft is collaborating with OpenAI on AGI development, and Open a OpenAI has obviously been the leader in the pursuit of AGI or one of the leaders. Alright. Say they're probably some of the first. Okay? And, Microsoft sees AGI as the next step in AI evolution, and they're actively researching AGI technologies. Y'all, it's 2 of the 3 largest companies in the world. Although they're not you know, that's not their mission.
Jordan Wilson [00:23:54]:
Achieving AGI is not their mission. They are supporting the very companies that do, and they are collaborating, and they are researching it. Meta. Alright. Here we go. Top 6 company in the, in the US. Meta, formerly known as Facebook. Right?
Jordan Wilson [00:24:08]:
They are actively working on AGI. They have
Jordan Wilson [00:24:12]:
a new project just announced to build an open source AGI, and their CEO, Mark Zuckerberg, views AGI as a key company goal. Right? He's really shifted his focus from you know, it was social media a while ago, you know, during Facebook's early days, then it was the metaverse. And now there's been a very hard pivot over the last year and a half toward not just artificial intelligence, but artificial general intelligence. K? Yes. He the CEO of one of the largest and most powerful companies in the world said achieving AGI is a key for his company. Would have been blasphemous 10 years ago, y'all. Anthropic. Alright.
Jordan Wilson [00:25:00]:
So Anthropic believes that AGI is imminent within a couple of years. They estimated human level AGI by 2030, and even the chief of staff this made big headlines a while back. Even the chief of staff believes that AI might take her job within 3 years. Yeah. The chief of staff, the person in charge of staffing, one of the most, influential and powerful companies in the AI space said that. Then we obviously have open AI. They've been actively pursuing AGI since their inception. That is part of their mission, is is achieving a safe AGI.
Jordan Wilson [00:25:37]:
So they believe AGI is inevitable and desirable, and they aim to develop AGI that outperforms humans on most tasks. And kind of their definition of it is a highly autonomous system that outperforms humans at most economically, valuable work. Alright. More receipts y'all. Let's talk about this. We're gonna take a little bit of a historical view here and talk about why these AGI predictions keep changing. Right? So I just mentioned there, Anthropic, OpenAI, Meta, NVIDIA. Some of the either CEOs or leaders of these companies are saying, we're going to see AGI in a couple of years, maybe 5 years, but definitely by the end of the decade.
Jordan Wilson [00:26:30]:
Which if I'm being honest, can be a scary concept. Right? Because it very quickly reshapes, AGI does, reshapes what humans are capable of and what machines are capable of. And that obviously has, both swift and long standing impacts to society, to business,
Jordan Wilson [00:26:56]:
to our worlds. Right? I'm not out
Jordan Wilson [00:27:00]:
here, you know, speaking in hyperbole. That's the truth. So let's talk about why these AGI predictions keep moving. Alright. So a very, kind of famous graph here, y'all. And, hey, for our podcast audience, this is one of the ones where you're gonna wanna check the show notes, and and and come back and maybe watch this on on YouTube or LinkedIn. And you can leave a question too, and I'll do my best to answer it or tag someone that can. Alright.
Jordan Wilson [00:27:31]:
But this is one of
Jordan Wilson [00:27:32]:
those charts I think you have to see, But
Jordan Wilson [00:27:35]:
I'm gonna do my best to describe it. So this is a chart from ARK Invest. Okay. And the title of this chart is expected years until a general artificial intelligence system becomes available. Alright. So, essentially, these are predictions charted over time on when leading experts, so these are averages, essentially how long until leading experts say that we will achieve AGI. Okay. So if you look at before GPT 3, so like I said, the GPT 3 technology was actually introduced in 2020 and made available.
Jordan Wilson [00:28:16]:
Right? Not I'd say most of the world didn't know about this technology until chat GPT in 2022. So if we look at pre 2020, the average expert said it would be at least 80 years. Okay? Let me repeat this. This is 5 years ago. The average expert said it would be 80 years until we achieved AGI. Okay? And then this graph here from ARK Invest kind of plots different, different key milestones and then how those milestones seemingly impact these experts. Right? Because this is, ARK Invest, I believe, was doing these studies on an ongoing basis and plotting them on an ongoing basis year by year. Or it looks like actually multiple times a year.
Jordan Wilson [00:29:12]:
Alright. So then we look when GPT 3 was announced, that 80 years went to 50 years. Okay? And then Google kind of got
Jordan Wilson [00:29:21]:
on board, right, with its Lambda models. Alright? And then it
Jordan Wilson [00:29:26]:
went from 34 years to average. Oh, how long is it gonna take? Went from 34 years to 18 years. Alright. Then chat GPT launched shortly. They're followed by GPT 4 launch, the premium and more capable version. Now and, you you know, I'm I'm gonna be interested to see, the next time this chart is updated. So, essentially, in late 2023, now we're at 8 years. So, again, post chat GPT, post GPT 4, now the average expert, kind of forecast for when we will officially achieve AGI is now a little under 8
Jordan Wilson [00:30:16]:
years. K?
Jordan Wilson [00:30:18]:
So again, let's talk about this. In 5 years, y'all. In 5 years, the smartest people in the world, the the biggest, brightest experts,
Jordan Wilson [00:30:28]:
in 5 years, they went from saying, we are 80 years from AGI to now, we are 8 years. Alright? And if I were to guess,
Jordan Wilson [00:30:42]:
I would say within a year, this chart will the average will be 3 years or less. Alright. And then you kind of see, you know, if this forecast error continues, that would plot this at about the end of 2026 or 2027, early 2027, so in, less than 3 years. Or if the forecast continue as they are now, essentially, it's the end
Jordan Wilson [00:31:12]:
of the decade. Right? We bring receipts here, y'all. We bring receipts. But I will also say so many experts don't understand
Jordan Wilson [00:31:25]:
AI. They don't understand generative AI. They don't understand large language models. They are, you know, future casting and and futurists and, you know, they're they're they're talking about, in theory, I wouldn't well, maybe at this point, I'm I'm an expert. Right? I put out 100 of of of podcast live streams, thousands of hours of contents talking to some of the smartest people, in AI in the world and and bringing them to you all too as well where you can ask them questions and learn from them. I'm surprised sometime. Right? When I read reports or I I read people who who write these, papers on on AGI and and AI, and I'm like, these people have no clue what they're talking about. I think even the average, quote, unquote, expert is disastrously misinformed or ill informed.
Jordan Wilson [00:32:18]:
And I think about AGI and and AI development, and I think this chart proves that. This chart proves that that the, quote, unquote, smartest experts in the world were laughably underestimating the power of artificial intelligence and the pace of its development. I think it is going faster than we understand. Alright, y'all? And, hey, if you do if you do have questions, let me know. John's asking if you think we can get, Sam or Jensen on the show, but we'll see. Maybe in the future here. We'll see. That would be great.
Jordan Wilson [00:32:58]:
Alright. Yeah, Brian. Same. Yeah. If I see one more AI bubble article. I did a full rant on AI is not in a bubble. Alright? So let's keep going here. We're not gonna keep this one going forever.
Jordan Wilson [00:33:13]:
This isn't gonna be an accidental 2 hour episode, but I wanna talk again a little bit more about this concept of the goalposts are constantly moving, and you know we bring receipts y'all. So I went back in the archives. Right? So I was reading articles from as far back as as Google caches could could find. Right? So I was reading articles from 30 years ago. Yeah. Some of the earlier ones. The website design was good. You you know? So reading articles from 30 years ago on what is AGI,
Jordan Wilson [00:33:48]:
there weren't a lot
Jordan Wilson [00:33:49]:
of them. But I read many of them from 30 years ago, 20 years ago, 10 years ago because I wanted to see something. I wanted to see something. Has the very definition of what artificial general intelligence is, has it changed? Right? Has it changed as AI becomes more powerful? Are we intentionally moving the goalpost back? Maybe because, I don't know, maybe because once you achieve AGI, it just makes things weird. So instead of saying, oh, okay. Hey. It looks by, like, by most definitions, maybe we're there. Instead of saying that, we just keep moving the goalposts and saying, oh, okay.
Jordan Wilson [00:34:26]:
Well, maybe it means something else. Alright. So this was from this was from the Machine Intelligence Research Institute, and this article was from about 10 years ago. Alright. And here's essentially a shortened version of what their definition was 10 years ago. So I'm summarizing here, but it said artificial general intelligence is the ability of a system to learn and solve problems across different areas, much like a human. AGI can achieve complex goals in various situations while using limited resources. AGI can also apply knowledge from one domain to another rather than just being good at specific tasks.
Jordan Wilson [00:35:19]:
Weird. I don't know. At least according to this article, which it seems like was one of the most prominent, you know, articles or pieces of research around AGI 10 years ago. I don't know. If I'm looking
Jordan Wilson [00:35:34]:
at these bullet points, I'm like, yeah. We're there. Right? Yeah. We are. Alright. Here's another one. This one was from about 11 years ago. Sam Altman.
Jordan Wilson [00:35:47]:
Maybe you've heard of him.
Jordan Wilson [00:35:49]:
He used to blog a lot. Alright. So this this article, Sam Altman essentially said alright. I'm paraphrasing here, but we'll put it in the newsletter. You can go read it for yourself. But he essentially, kind of defined artificial intelligence, as something that possesses the ability to understand, learn, and apply knowledge across a wide range of tasks similar to human cognitive abilities.
Jordan Wilson [00:37:16]:
Uh-huh. Aren't we kinda there? Right? So the goalposts keep moving.
Jordan Wilson [00:37:21]:
So it's like, have we achieved AGI? I I don't know. Seems like we keep changing what AGI is because at least according to some of these things from 10 years ago, right, you can look at these and be like, yeah. We we would yeah. We're here. Right? Obviously, as, researchers learn more, they start to set maybe more detailed and specific, I guess, boundaries or milestones on what achieving AGI even is and what it means. Sure. I get that. You know, 10, 20 years ago, I I I think AGI was actually, for the most part, a little more theoretical.
Jordan Wilson [00:38:01]:
Right? And then as we actually get closer and closer to potentially achieving artificial general intelligence, then we start to set stricter guidelines on what are the hurdles to clear in order to officially say that we've been there. Alright. We got 2 2 more points here y'all, and these are big ones.
Jordan Wilson [00:38:21]:
So here is, I think, one of the kind of secret reasons that maybe why we haven't technically achieved AGI. And I think it's actually this OpenAI and Microsoft partnership.
Jordan Wilson [00:38:40]:
Okay? So I'll give you this super super high level overview here. So this partnership between Microsoft and OpenAI, reportedly, Microsoft invested about $13,000,000,000 and has a 49% stake in 49% ownership stake in
Jordan Wilson [00:38:59]:
OpenAI. However, once the OpenAI
Jordan Wilson [00:39:04]:
board says that it has achieved AGI, that agreement changes. Alright? And keep in mind that Microsoft did previously have a seat on the board at OpenAI. Right? Interesting there. It no longer does. Alright. So interesting. Right? Let's let's let's keep let's keep going. So, essentially, Microsoft doesn't have any stake in OpenAI's future AGI Tech.
Jordan Wilson [00:39:38]:
Right? So, again, I'm sure they have, legal documents, right, that are, I'm sure 100 of pages long that aren't publicly available, but what is publicly available is essentially future AGI technology from OpenAI. Microsoft does not have a stake in that. Right? So that is when their current partnership could start to change. And I guess it obviously depends on what's in those, documents that the rest of
Jordan Wilson [00:40:10]:
the public does not really have access to. But my question is, how do they separate it? Because my
Jordan Wilson [00:40:20]:
outsider's viewpoint, right, I would consider myself a very informed, outsider, but I'm an outsider nonetheless. So take my hot take here on hot take Tuesday with a grain of salt. Right? But I feel, you you know, and I believe this, investment was made about 5 years ago.
Jordan Wilson [00:40:39]:
I think at the time neither company could foresee
Jordan Wilson [00:40:45]:
what this partnership would mean and how big artificial intelligence and large language models, how pivotal they would become in our day to day operations. Right? I guess maybe best case scenario when Microsoft made this large investment, maybe they said this is going to power, power the future of our operating system. Right? Maybe they said best case scenario. Maybe they were just taking out an expensive flyer on very promising technology. But think now, if you're a power Windows user, you're probably using Microsoft 365 Copilot. Right? Maybe your organization is dependent on this Copilot technology that is baked into the operating system of Windows, which is, guess what, powered by OpenAI's GPT 4 o technology. So I don't know if when this agreement was originally staked, if either company could have seen into the future and could fully understand how important this partnership would be to not just both of their respective companies, not just to the business world, but to the US economy. I don't think people understand, and I've gone over this in-depth.
Jordan Wilson [00:42:09]:
I'm not gonna go over it again. Right? Essentially, I I said, hey. If if if if you're a knowledge worker out there in the US, you have no clue, but you are constantly using OpenAI's technology. You just don't know it. In your daily lives, in your personal lives, in your business lives, your company, the software you use, everything, whether you know it or not, is somehow powered probably by OpenAI. But, also, if OpenAI, if their board says, yes, we have achieved AGI, we have the technology,
Jordan Wilson [00:42:43]:
how do they separate it? Right? OpenAI made a pretty big and, I think, a smart move when they came out with GPT 4 o, and o means omni, where essentially everything is being done under the hood by a single model. Right? And, presumably, OpenAI is working very hard to make that singular model more and more powerful, and to give it, in theory, capabilities that would reflect artificial general intelligence. So how
Jordan Wilson [00:43:16]:
do they separate it? Right? We know that the partnership is both as important, I would guess. It is equally as important to Microsoft and to OpenAI. So what would be OpenAI's
Jordan Wilson [00:43:33]:
reason to come
Jordan Wilson [00:43:34]:
out and say, yes. We have achieved AGI. Therefore, we must somehow restructure our technology. We must separate our technology. We must dilute somehow or start to dissolve or rework our partnership with Microsoft. What incentive do they have to to say that to do that? Right? And, also, OpenAI, achieving AGI has been mission critical to them, and they've raised 1,000,000,000 of dollars. I don't know. Maybe I'm speculating here because it's hot take Tuesday.
Jordan Wilson [00:44:08]:
But if I'm an investor looking to potentially invest 100 of 1,000,000 or 1,000,000,000 of dollars into OpenAI, right, there's been reports that Sam Ullman is looking to raise $7,000,000,000,000, for some future projects centered around intelligence, centered around, compute power, etcetera. I don't know. If said company OpenAI has, quote, unquote, achieved its mission, if one of its founding missions is to achieve AGI and the board comes out and says, yes. We achieved AGI. Doesn't that make fundraising exponentially harder for Sam Altman and OpenAI? I'd say yes. So what incentive then? Right? Because if OpenAI comes out and says, yes. We've achieved AGI. It's great.
Jordan Wilson [00:44:50]:
Right? Big box checked off. Huge step for for humanity. Huge step for technology. Huge step for the future of business. What does OpenAI actually have to gain? I don't know. I'd say they might have more to lose to say that they've officially achieved AGI. Right? So I think and I wouldn't blame them. I would say OpenAI OpenAI continues to move the goalpost because OpenAI, I think, also loses in some way, shape, or form again.
Jordan Wilson [00:45:21]:
I could be ill informed because I don't have access to the documents and, you know, the public doesn't as either. But it is clear that the partnership changes. The OpenAI Microsoft partnership changes. So, yes, OpenAI has greatly benefited from the funding and the support and the architecture that, Microsoft has provided. But at the same time, I think OpenAI is benefiting in a huge way as well. If I if if I'm an OpenAI executive, I don't wanna lose that partnership. Think of how much insights they have to gain, right, when they are presumably getting reports back, from Copilot users to improve the GPT 4 technology. That's some of the most important information from a startup is essentially when us humans are telling them what's a good output and what's not.
Jordan Wilson [00:46:09]:
That saves them years of development time and probably 1,000,000,000 of dollars of development time in the end. K? So we have to think about that. Alright. Let's wrap this thing up, shall we? So why are we closer than ever? Why are we closer than ever to AGI? Well, like I talked about. It has been unprecedented up until this point that some of the largest companies in the United States are openly working toward achieving AGI or actively supporting and openly supporting those companies that are doing that. Like I said, a decade ago, more you
Jordan Wilson [00:46:58]:
you know, 15, 20 years ago, it would have been straight up taboo. As a big company, as a public company to say, yeah. We're working toward AGI. People would have called you
Jordan Wilson [00:47:08]:
a apocalyptic crazy, and and and and your stock, if you were a public company, would have went in the toilet. Now now if you're a big tech company like Amazon, like like Meta, like Microsoft, you almost have to either be explicitly and outwardly working toward AGI or indirectly how you know, working there. Because I think the dollars have started to make sense of what AGI means to business and to the US economy. Someone write that down. That was good. Right? Yeah. This is on y'all. This is unscripted and unedited.
Jordan Wilson [00:47:48]:
Aside from for some from some, from some bullet points I put up on the screen. Alright. So that's number 1. Also, 2 of the most powerful a a AI startups in the world are either directly working toward it or indirectly supporting it in OpenAI and Anthropic. And then last but not least y'all, and this was kind of, kind of referenced at the beginning of the show when we talked about, Google Gemini giving away 1,500,000,000 tokens per day for
Jordan Wilson [00:48:19]:
developers. That's
Jordan Wilson [00:48:22]:
that's wild. That's wild. Right? And then also OpenAI similarly said, hey. GPT 4 o Mini, it is free. They announced this more, more than a month ago. They said it is free to fine tune through essentially the end of September. So these companies have essentially been giving compute. They've been giving tech like, this this, technology, this this inference power.
Jordan Wilson [00:48:54]:
They've been giving away intelligence for free. Right? It is now to the point. Right? There's this saying in the AI community. Intelligence, too cheap to meter. One of the biggest obstacles to achieving AGI 5, 10, 15, 20 years ago was the cost. It was the cost and the availability. If you were a tech company 10 years ago or if you were a a scrappy startup 10 years ago, and if you wanted to work toward AGI, it would be nearly impossible. It would be nearly impossible.
Jordan Wilson [00:49:31]:
Right? Now, not so much. Right now, if you want to give AGI esque capabilities to your business, right, It is essentially free to do right now. You y'all, like 10, 15, 20 years ago, you had to have 1,000,000. 1,000,000, or 1,000,000,000 of
Jordan Wilson [00:49:51]:
dollars. It's free now.
Jordan Wilson [00:49:55]:
Right? For businesses, at least. Right? Obviously, these companies, this is a a cost for them to bring people into their platforms and to keep them there, but there's a race for this. But, also, even for these companies, The cost of compute has gone down exponentially. Right? Essentially, GPU chips. So, these these chips that these big companies need to create next generation AI. 10 years ago, the chips used to be much more expensive and much less powerful. The cost of compute is going down exponentially. Most estimates say even in
Jordan Wilson [00:50:34]:
the last couple of years, it's it's gone down tenfold. Tenfold.
Jordan Wilson [00:50:40]:
The same thing with cloud computing. Tenfold in years. So the rate at which intelligence and GPUs and the cost to get to AGI has gone down exponentially in the last. Jeez. Just the last couple of months, it has turned into a sprint into giving companies essentially intelligence, giving them compute for free. Alright. We are in unprecedented times, at least when it comes to AI development, and if and when AGI is possible. Alright.
Jordan Wilson [00:51:22]:
I'll wrap it up by saying this y'all. I'll wrap it up by saying this. I think we're gonna be there very soon. I think we are going to achieve AGI very soon. I just gave you all the receipts. If we're looking at how AGI was commonly defined 10, 15, 20 years ago, we've already achieved it. The definition of today, I think will be there in less than 5 years. It's a given.
Jordan Wilson [00:51:59]:
But will that definition continue to move? I would say so. But y'all, whether you are just an individual who's interested in AI, a business leader, or maybe you are someone who's working, in the AI space. AGI is inevitable. Right? You saw on that chart 5 years ago, they said we were 50 years out, or they said no. Sorry. They said we were 80 years out 5 years ago. Now they're saying 8. We've gone from an 80 year projection to an 8 year projection.
Jordan Wilson [00:52:37]:
And like I said, I think when the next time these charts are updated, it's gonna be like 3 years. We are going to get there very soon. Alright? So what does that mean for the future of work? What does that mean for the future of business? What does that mean for your future career? I don't have those answers yet, but that's why we're going to be here every day helping you figure that out, bringing on experts from across the world, from all these big tech companies. We're giving you examples of how companies large and small are growing with generative AI, how people are changing the trajectory of their careers, real use cases. That is what everyday AI is all about. Alright? So when will we achieve AGI? I'd say sooner than we might think.
Jordan Wilson [00:53:28]:
Like I said, by old definitions, I think we're already there, but I think it's going to be a matter of years. Alright. I hope this
Jordan Wilson [00:53:36]:
is helpful y'all. If so, please repost this. Share this with someone in your network. I don't know. Maybe if you do, I'll send you even more spicy takes that were too spicy, for the show. Alright. So, if this was helpful, please let us know. If you're listening on the podcast, please, click that follow button.
Jordan Wilson [00:53:52]:
Leave us a rating if this is helpful. We put so much time, energy, and effort, into cutting through all the nonsense and giving you what actually matters. So also make sure if you haven't already, go to your everydayai.com. Sign up for our free daily newsletter and check out the thanks a million giveaway to celebrate, the everyday AI podcast hitting a million downloads. We are going to be giving away a lot of cool prizes, giveaways, a year of, you know, chat gbt or your favorite large language model for whoever has the most referrals. So when you sign up, you're going to get a unique URL. Then you can send that to anyone. You can post that online.
Jordan Wilson [00:54:29]:
You can email that to someone. You can text it to them, whatever. But if this show is helpful, if the newsletter is helpful, if the live stream is helpful, if if getting access to world leading guests and being able to ask them questions is helpful, please share this with your friends. Also, please join us tomorrow and every day for more everyday AI. Thanks, y'all.