Understanding the milestones of Artificial Intelligence (AI). Today’s story is centered around the transformation of AI. It involves the ...
 |
| Understanding the milestones of Artificial Intelligence (AI). |
They are betting not only on money but also on the transformation of our jobs and a paradigm shift in the social structure as we know it today. This exciting and unsettling future is in the hands of a small group of people. We could mention the well-known Sam Altman, CEO of OpenAI, who, as we know, has had incidents that have caused the entire sector to come to a standstill.
It was after one of those crises of confidence that what we now know as Anthropic was born. Likewise, a more recent event nearly caused the complete collapse of OpenAI. Another key player is Elon Musk, who, after distancing himself from Larry Page (founder of Google), paved the way for the creation of OpenAI.
We could also mention Mark Zuckerberg, Satya Nadella, or Sundar Pichai. And let’s not forget how Peter Thiel (founder of PayPal) kicked off the race for AI talent back in 2012. These individuals, who are shaping the future of our lives, have visions that align for a time, only to eventually lead to distrust and rivalry.
Before AI
In the 1950s, researchers began exploring the possibility of “teaching” machines to respond to human questions. They identified a critical milestone in the development of this technology, which they called Artificial General Intelligence (AGI). AGI would be achieved when machines could perform tasks equivalent to those of the human brain, such as learning to speak, think, acquire knowledge in various fields, and make autonomous decisions.
This idea remained stagnant for decades due to the lack of significant advancements in its development. However, this changed with the arrival of two British scientists, Demis Hassabis and Mustafa Suleyman. Convinced that 2010 was the right time to initiate a plan to achieve AGI, they considered factors such as advances in computational processing and the internet, which had made a wealth of digital knowledge available to feed intelligent models.
During this period, they met Eliezer Yudkowsky, one of the early proponents of the idea that AI could solve some of humanity’s greatest problems, such as cancer and poverty. Yudkowsky successfully built a community of theorists, academics, and investors interested in AGI. Among these community members was Peter Thiel.
Hassabis, Suleyman, Yudkowsky, and Thiel formed a relationship centered around the idea of creating technology capable of revolutionizing the modern world. This led to the founding of DeepMind in 2010, a project that marked the beginning of modern AI. Hassabis and Suleyman’s strategy to launch the AGI race was to teach machines to beat humans in video games.
Some might argue that IBM was a pioneer in this approach by using chess as a learning medium, culminating in Deep Blue’s victory over Garry Kasparov in 1997. However, due to the high development costs, IBM only used this technology as a marketing strategy, without fully exploring its possibilities. With funding and recommendations from Peter Thiel, Hassabis was invited to SpaceX headquarters for a meeting with Elon Musk. Following the meeting, Musk decided to invest in DeepMind.
Subsequently, Hassabis and Suleyman convinced Larry Page to join their project, resulting in DeepMind gaining significant investors and allies in Thiel, Musk, and Page. This development did not go unnoticed in Silicon Valley, sparking interest from other tech giants like Microsoft, Baidu, and Meta. This raised the question of where to find individuals with the expertise to fully understand AI’s potential. But where could they find such talent?
Human Talent in the AI Field
In Toronto, Geoffrey Hinton, along with two graduate students, including Ilya Sutskever, presented a research paper titled “The Revolutionary Technique That Quietly Changed Machine Vision Forever.” They developed a neural network capable of identifying the content of an image with remarkable accuracy. This breakthrough was largely due to an initial investment of $12 million from NVIDIA, which was primarily focused on selling video game graphics cards at the time.
Recognizing the potential of his project and team, Hinton sought additional financial backing to ensure its continued progress. Google recognized this potential and decided to integrate them into its advanced models research center. With this development, Hassabis and Suleyman realized the challenges they would face in retaining their talent at DeepMind.
How could they compete with offers from big tech companies? Understanding this, they decided to sell DeepMind, with conditions that the technology would not be used for military purposes and that all activities would be overseen by an independent board. Google once again took the lead, acquiring the pioneering AI company. This acquisition, along with the addition of the research team that had made significant advances in modern AI, suggested that Google was monopolizing top talent. Besides DeepMind, Google also had Google Brain, another division dedicated to AI.
The Beginnings of OpenAI
One of the key figures in modern AI is Sam Altman. Back in 2015, he was the CEO of Y Combinator. He connected with Elon Musk, and they quickly found common ground on their goals for advancing AI. They knew that to build technology with the impact they wanted, they needed to get other researchers on board, which they did. They also received support from Musk’s former partners, such as Peter Thiel and Reid Hoffman (founder of LinkedIn).
Besides their main goals, this new group wanted to counter Google’s growing monopoly in AI. On December 15, 2015, they publicly announced the creation of OpenAI. Musk made an initial investment of $40 million to get things started. OpenAI was set up as a non-profit, dedicating all its resources to research and technology to achieve AGI.
They released their model as open source so others could use its full potential at no cost. (It was the failure to stick to this open-source commitment that later led Musk to leave the OpenAI project.) It’s also worth mentioning that Musk was a major driving force behind OpenAI and was instrumental in convincing Ilya Sutskever to leave Google and join their team.
The Major Hidden Change
It was 2016. There wasn’t any significant AI product launch that year, but AI was gradually capturing the world’s interest. Here’s the paradox, though: that same year saw the publication of one of the most impactful AI research papers ever — “Attention Is All You Need.” This might be the most important discovery in AI in the past 50 years. Without this research, we probably wouldn’t be talking about ChatGPT, Gemini, Claude, and more today. If we think about it a bit more, what we know about NVIDIA and its current boom would be unimaginable without this development.
That’s how significant it is. The paper introduces the concept of the transformer, a new way of teaching machines. This mechanism for language models allows them to translate texts more quickly, among other things. The basic idea behind transformers is converting text into numerical components called tokens.
This change exponentially improved the learning capacity of traditional models and, at the same time, required high computational processing power. This naturally limited access for various companies. That’s where NVIDIA began to gain the relevance it holds today.
Final Thoughts
I believe that, like any other revolutionary historical event, the development of AI as we know it today involved a series of decisions and risks. Initially, Google implemented strategies that were expected and even anticipated. However, as we’ve seen, it took more than just money; it required disruptive ideas from a multidisciplinary team with the talent needed to push AI development forward.
Today, it’s easy to overlook the details we’ve explored together in this article. But it’s crucial to understand these details to grasp the initial goals set around AI and to see the transformations it has undergone, for better or worse.