Artificial Intelligence. AI. By now, you’re probably sick to death of hearing how it’s the next big thing. This not-so-new concept has taken over the tech industry, and it was always only a matter of time. See, artificial intelligence has been around since the ’50s, when the concepts of machine learning and the capabilities of computers to think were first explored. However, aside from a few board game protegees and helpful yet simple-minded assistants (I’m talking to you, Siri), the potential of this research would remain mostly untapped, with the theory far ahead of the tech. The ‘modern AI’ we know and dread, really materialised in the last 10 years, spearheaded by OpenAI.
A brief history of OpenAI
Founded way back in 2015 as an open-source artificial intelligence organisation, OpenAI’s mission was to advance the field of AI in an ethical and socially responsible manner. The company were quick out of the gate with an initial $1 billion startup investment. Using the deep learning concept involving the creation of neural networks to mimic how the human brain learns, they began the development of AI models capable of synthesizing information and fulfilling tasks from basic prompts. In 2019, the organisation controversially restructured from a non-profit to a ‘capped-profit’ model, due to the financially demanding nature of their research, and a realisation of the staggering earnings potential I’m sure. Not long after, Microsoft came knocking with another billion dollars, to get in on what was a very promising startup.
ChatGPT
Following the development of GPT-3, OpenAI’s first Large Language Model (LLM), AI was ready for the public (whether the public was ready for it is another question). At the end of 2022, they released ChatGPT, a chatbot with capabilities unlike anything seen before, which became the fastest-ever app to 100 million downloads, reaching this milestone in just two short months. Though far from the finished product, this chatbot could process information and generate fast and (mostly) accurate responses, blowing minds around the world.
At the beginning of 2023, right when people were getting their heads around the incredible capabilities of Open AI’s chatbot, and as educators and other concerned parties were figuring out how to detect AI-generated content, they go and release GPT-4, 10 times as powerful as its predecessor, far better at producing original content and understanding more complex prompts. Additionally, the new DALL-E model was showing significant improvements in image quality generated. Essentially, this model was smarter, and it became apparent that the growth trajectory of this tech was not going to be linear. It was going to be exponential.
AI Competition
Given the extreme popularity of AI amongst the general public, all the big fish in tech are dipping their toe in. Samsung’s new range of phones, the S24 range, comes with their new Galaxy AI technology. Microsoft, along with investing in OpenAI, has its own Copilot, which integrates with its existing products. Amazon is exploring similar avenues with their cloud-computing services. Oh, and Google, with their DeepMind subsidiary, released their new multi-modal Gemini 1.5 AI model earlier this month, with more processing power than anything seen previously, including GPT-4. All these multi-billion dollar corporations see the reality of AI, as a market which, by most estimations, will hit the trillion dollar valuation by the end of this decade. Make no mistake, there is an arms race going on right now, and it remains to be seen whether ethical and socially responsible choices will be prioritised as these companies attempt to carve themselves out a slice of this emerging market.
Sora
So now that we’re up to speed on AI and particularly OpenAI, let’s look at the most recent development in the field, which has put the AI industry on notice. On February 15th, OpenAI released footage of Sora, a video-generation model which many thought was years away from being achievable. Naturally, given the fact that this is a video-generation model, I can’t really demonstrate just how f**king mental it is, so go have a look for yourself, if you haven’t seen already.
It’s pretty incredible. For context, this time last year, this is where AI video generation was.
Sora, described by OpenAI as a world simulator, has shown the ability to learn and implement the laws of physics and is a visual demonstration of the power of AI. This considerable leap forward is of course causing speculation about how long it will take before we reach the Holy Grail of artificial intelligence.
The AGI Rabbithole
Artificial General Intelligence (AGI) describes an AI model which surpasses anything we have currently. A model with the ability to adapt without interference, capable of understanding topics without training, apply logic, think independently, and act with autonomy, essentially possessing human-level intelligence. Such a model seemed to be a far-away aspiration for AI research, however, recent innovations, and especially the rate at which these innovations are being made, have consistently cut the timeframe in which such tech may be possible.
Remember Thanos? That big purple guy with a passion for stones. Well, just like Thanos, OpenAI seems to be filling up its gauntlet with all the pieces needed for this revolutionary, potentially destructive creation. They have already managed to produce an AI that can speak, hear and see (ChatGPT). Their model also has internet browsing capabilities, and as such, virtually unlimited information access. Sora, their newest toy, has shown an ability to use logic, predict motion, and dynamically reason with and understand physical environments. Oh, and yes, they are working on AI that can perform tasks unsupervised, by accessing your computer for example, and carrying out actions without prompts. There’s your autonomy. What you have coming together is a collection of qualities that begin to resemble human intelligence. Maybe not all the way there, but dramatically closer than would have been expected a few years ago. There’s one problem though. To assemble an AGI model with all these qualities combined would likely take a lot of resources. A LOT. The scale of which could not be met by the current AI infrastructure.
But wait. Just last week, Sam Altman, CEO of OpenAI, casually mentioned that he would be attempting to raise $7 trillion to invest in GPUs, the hardware needed for building AI models. And no, that’s not a typo, he really said SEVEN TRILLION. As in, 7 million million. I didn’t even know you could ask for that. Anyway, this has naturally been subject to a lot of scrutiny in the media, with some suggesting raising such a sum would not be possible. Yet, it demonstrates a vision of aggressive growth in their operational capacity, and this may be the last piece of the puzzle for the lofty goal of AGI creation. It’s all speculation of course, but given how quickly developments are occurring in AI, it is a compelling theory.
An AI Future
Pandora’s Box
noun
1. a prolific source of troubles
This leads nicely into an important discussion that needs to be had about AI. This technology has arrived, and will probably be the most important invention of our lifetime, but how we use it will be just as crucial to our collective outcome. It brings with it the potential to better the human race, with democratised access to information and the ability to solve problems around health, education and the environment, pushing scientific advancements to previously inconceivable places. But it also poses risks. The improper use of AI could lead to mass misinformation, mass replacement of jobs, and if you really want to go there, sentient AI entities which may have little regard for us pesky humans as they ascend far beyond our levels of intelligence. Call me an optimist, but I prefer to see the positive potential of AI. Certainly, there is much to be concerned about during such drastic change. Parallels can be drawn with the advent of the internet. Purely by human nature, there will be a ‘dark web’ type element to AI. It’s unavoidable. But as long as there is enough good to outweigh the bad, it is surely worth exploring.
An example of the good, and get ready to have your mind blown:
This entire article was actually written by AI.
No, not really. If that were the case, I’d be managing to get more than one of these out a month. It’s another interesting question though. Is there an intrinsic, irreplaceable quality to human creativity, or will AI be better at that too? I sure hope not, but nothing is really out of the question when AGI inevitably arrives. Will there just be a limitless amount of content created for and tailored specifically to you, that never runs out? And is that a world we want to live in?
One way or another…
AI is revolutionary, in the same way that the wheel, printing press, and telegram were. In the same way the internet was. The difference is, that those technologies were developed and implemented over decades, centuries even. In today’s globalised, interconnected, instant world, new tech goes out to everyone, everywhere, all at once. AI is already used by millions, and it’s only been a couple of years. It was thought we would see some form of AGI during our lifetime, however, with each new innovation, it’s looking more like during this decade. Is humanity ready for such a powerful tool? Is ‘tool’ even the correct term for something which surpasses any intelligence we have seen to this point? Watch this space. In the coming years, this technology is going to reach new unforeseen heights and infiltrate every aspect of our lives. Sure, for now, it’s just cool videos and homework helpers, but if OpenAI has shown us anything, it’s that change is quick in the AI world. In a few years, when we’re all strolling around like Tony Stark, talking to our genius personal assistants located in a chip stuck onto our brains, we may cast our minds back to this time as the beginning of a new age. Or maybe we’ll be enslaved or dead, courtesy of our robot overlords. Who knows?