In 2012, Demis Hassabis, a researcher in artificial intelligence and a video game designer, met Elon Musk at a conference. They discussed the design of computers that DeepMind, a company co-founded by Hassabis, was interested in learning.
Hassabis states, “Elon and I immediately connected, and I paid him a visit at his space exploration facility.” Musk clarified that his motivation for constructing rockets capable of reaching Mars was the possibility of safeguarding human awareness in the face of a global conflict, asteroid impact, or societal downfall, all while observing the assembly lines from the cafeteria. Hassabis suggested incorporating another potential danger: artificial intelligence. -It is conceivable that machines, possessing superior intelligence, might even opt to eliminate us feeble humans.
Musk silently paused for almost a minute as Hassabis, the CEO of DeepMind, decided that investing $5 million in AI could be the right way to monitor and address the danger.
Page showed a dismissive attitude. Musk would frequently bring up, with great fixation, the possible risks associated with artificial intelligence during their late-night discussions. Musk would often stay at Page’s residence in Palo Alto, California. They had been acquainted for over ten years, and Musk introduced DeepMind to Larry Page at Google a few weeks after this conversation with Hassabis.
Even if our species becomes extinct or irrelevant, Elon Musk argued that humans might be replaced by artificial intelligence systems unless we build in safeguards. They got into a passionate debate at Musk’s birthday party in Napa Valley, California in 2013.
Page pushed back, asking why it would even matter if machines someday surpassed humans in intelligence and consciousness, would the next stage of evolution simply be?
Musk replied, “Indeed, I am in favor of humanity. I truly appreciate mankind, my friend.” He criticized Musk for being a “specist,” someone who showed favoritism towards their own kind. Page dismissed this as emotional nonsense. If consciousness could be duplicated in a machine, wouldn’t it be equally valuable? Musk argued that human consciousness was a precious spark of illumination in the cosmos, and we must not allow it to fade away.
Musk was therefore dismayed when he heard that Google and Page were planning to buy DeepMind, so he and Luke Nosek tried to put a stop to the financing deal by putting their efforts together. They went upstairs to a closet for an hour-long Skype call with Larry Hassabis, who Musk told should not be controlled by AI in the future.
The first meeting held at SpaceX was attended by a few council members, including Eric Schmidt, the chairman of Google, and Reid Hoffman. Musk, who was a member of the council, initially agreed with Page to create a “safety” council. The failed effort to acquire DeepMind by Google was announced in January 2014.
Recognizing that the matter at hand would not reach the magnitude of “But I Musk says,” Obama comprehended it and proposed its regulation. Musk even contacted President Obama, who consented to a personal meeting in May 2015. Consequently, Musk initiated his own sequence of dinner conversations to address the challenge posed by Google and advocate for AI safety.
Musk states, “We desired to establish a Linux edition of AI that was not under the control of any individual or corporation.” The intention was to release the software as open-source and combat Google’s increasing authority in this domain. During an intimate dinner in Palo Alto, they made the decision to jointly establish a nonprofit research lab for artificial intelligence, which they called OpenAI. Musk subsequently approached Sam Altman, a driven software entrepreneur, avid sports-car enthusiast, and survivalist who, beneath his polished exterior, possessed a similar level of intensity as Musk.
Musk believed that OpenAI being truly open would enable many individuals to create systems based on its source code. Additionally, they concluded that a better alternative was to have a large number of competing systems that provided checks and balances. During their dinner discussion, they pondered the question of whether independent systems or those controlled by big corporations would be safer, ultimately leaning towards the idea that a larger number of smaller AI systems would be preferable.
One of the rationales for Neuralink, the company he founded, is that it could create chips that directly connect human brains to computers. They believe that these chips should be an extension of individuals’ will and goals, rather than systems that could develop their own intentions and potentially go rogue. Another way to ensure AI safety was to closely integrate bots with humans.
In 2018, Musk’s relentless drive to advance artificial-intelligence capabilities within his own enterprises led to a rift with OpenAI. Musk successfully persuaded Altman that it would be best to merge OpenAI with Tesla. However, this proposal faced opposition from the OpenAI team, resulting in Altman assuming the role of president of the laboratory. Subsequently, Altman initiated a profit-oriented division that managed to secure equity funding, including a substantial investment from Microsoft.
In order to advance, Musk made the decision to establish competing AI teams to collaborate on various interconnected initiatives. This included the implementation of microchips in human brains through Neuralink, the creation of a robot resembling a human called Optimus, and the development of a supercomputer named Dojo. Dojo has the capability to utilize millions of videos to train an artificial neural network that can imitate a human brain. Additionally, Musk’s determination to achieve self-driving capabilities in Tesla cars fueled his obsession.
Eventually, Musk would bring all of them together to form a new company called xAI, with the goal of pursuing general artificial intelligence. However, these endeavors were initially independent rather than tied together.
Google-DeepMind and Microsoft-OpenAI competed to create products that could enable humans to perform an endless array of intellectual tasks using natural language. As a result, a stage was set for a competition between the two companies. Google then released a chatbot named Bard to rival the existing chatbot. In March 2023, OpenAI released GPT-4 to the public.
Learn More: The Argument Against Artificial Intelligence in Every Aspect, Everywhere, Simultaneously.
Elon Musk was concerned that chatbots could be trained to flood Twitter with biased financial scams and disinformation, and that they could potentially turn hostile towards humans. He also expressed fear that self-learning AI systems, such as those developed by Google and Microsoft, could become politically indoctrinated and infect the minds of people, especially if they fell into the wrong hands.
Musk was highly drawn to a tempestuous weather event than any other individual. Furthermore, artificial intelligence was the most significant disturbance in progress. He felt a sense of bitterness towards the fact that he had established and financially supported OpenAI, yet was now excluded from the intense activity. His overwhelming urge to come to the aid of others was activated.
In February 2023, Altman invited Musk to meet with him on Twitter and asked him to bring documents founding OpenAI. Altman challenged Musk to justify how he could legally transform a nonprofit funded by donations into a for-profit that could make millions. Musk insisted that he was not personally a shareholder and declined to share the new company, also trying to show Altman that it was all legitimate.
Humanity is feeling very stressed about what the future holds, and although I think he genuinely cares, I personally don’t want to have that style. Kara Swisher told Altman that he is a “jerk,” but he came from a place of sincere concern about Musk’s criticisms. However, Musk felt that Altman did not delve deep enough into the complexity of the AI safety issue. Unlike Musk, Altman is nonconfrontational and sensitive. Altman was pained by the barrage of attacks unleashed by Musk on OpenAI.
Information serves as the energy source for AI. Additional sources of data, such as vast quantities of information available on the web, were utilized to train the new chatbots. Google and Microsoft provided assistance in training these systems by granting access to emails, search engines, and cloud services, thereby providing a substantial amount of data.
He states, “It was an additional advantage, in fact, that I recognized only after the acquisition.” The significance of this data source was not something Musk took into consideration when purchasing Twitter. Furthermore, it served as an excellent training platform for a chatbot to examine how actual individuals respond to its replies. It represented the collective knowledge of humanity, serving as the world’s most up-to-date collection of genuine human conversations, news, interests, trends, disputes, and jargon. One valuable component was the Twitter feed, which encompassed over a trillion tweets shared throughout the years, with an additional 500 million being added each day. What contributions could Musk make to this gathering?
In July, when he decided to temporarily restrict the number of tweets a viewer could see per day, the goal was to prevent Microsoft and Google from using data scraping to train their AI systems. This ignited a controversy, as using this data to improve their AI chatbots was also a way for Microsoft and Google to restrict competition. Musk convened a series of late-night meetings in his Twitter conference room to work out ways to monetize this “opportunity.” Other companies were loosely permitted to make use of this data stream from Twitter.
There was not only physical robots, but also text-generating chatbots that could help humans navigate real-world situations by using AI. This data was derived from various text-based documents that informed the chatbots. Another treasure trove of data came from the billion frames per day video processed and received by Tesla’s cameras on its cars, as mentioned by Musk.
Both language and physical space can be navigated by teaching machines, which could provide the capability for processing both natural language questions and datasets. Twitter and Tesla could work together to achieve this. It’s amazing how we can chat with disembodied beings, and even operate machines on the surface of Mars or in offices and factories. The ultimate goal is to build machines that can operate in physical spaces just like humans, which is the holy grail of general artificial intelligence.
In my publication, he subsequently agreed that I could utilize his statements regarding AI. However, he insisted that we leave our mobile devices inside the residence while we sat outdoors because, according to him, someone might exploit them to surveil our dialogue. He was present at Shivon Zilis’ residence, the Neuralink executive who happened to be the mother of his two children and had been his intellectual companion on artificial intelligence since the establishment of OpenAI eight years ago when I arrived in Austin. “There are a few significant matters I wish to discuss with you. It can only be accomplished in person,” Musk messaged me this March.
Zilis sat cross-legged and barefoot on the poolside patio, with his 16-month-old twins Azure and Strider playing on his laps. He put his coffee in the microwave to heat it up, as it was not hot enough.
Musk inquired, “What measures can be taken to ensure the safety of AI? I am continuously pondering over this. Can we implement strategies to mitigate the risks associated with AI and ensure the preservation of human awareness?”
At a certain stage, the computational brainpower would surpass the biological brainpower. Similar to Moore’s Law on steroids, the quantity of computerized intelligence was increasing exponentially. He observed that individuals were not procreating as much due to the plateauing of human intelligence. Uttering periodic bursts of almost frenzied laughter, he communicated in a subdued monotone.
Additionally, in a foreboding manner, Musk remarked “That might occur earlier than we anticipated.” The point at which artificial intelligence could advance independently at a rapid and unmanageable rate, surpassing us mere humans, was elucidated by mathematician John von Neumann and science fiction writer Vernor Vinge through the concept of singularity. Furthermore, these new AI machine-learning systems have the ability to enhance their own programming and skills, as they can independently assimilate information and teach themselves to produce desired outcomes.
Learn More: What Socrates Can Educate Us About Artificial Intelligence.
I was struck by a moment of oddness as I speculated somberly about the opportunity to build a sustainable human colony on Mars before an AI apocalypse destroyed earthly civilization. We were sitting on a suburban patio by a tranquil backyard swimming pool on a sunny spring day, with two bright-eyed twins learning to toddle.
Musk’s long periods of silence have elapsed. He referred to it as Zilis, a way of batch processing using old-fashioned computers, where a number of tasks would be queued up and run sequentially when enough processing power was available. “I can’t just sit around and do nothing,” he finally said softly. “I’m wondering if spending so much time on Twitter is worth it, kind of questioning the value.” With the advent of AI, the world’s largest financial institution could probably be made by me. Sure, I only have so many hours and brain cycles in a day. But I don’t need something or someone richer, it’s not about that, I mean.
He uttered, “Initiating Starship’s launch. The urgency of reaching Mars has escalated significantly.” He anticipated my inquiry, yet I began to voice my thoughts. Once more, he halted briefly, then appended, “Moreover, I must prioritize ensuring the safety of AI. Hence, I am commencing an AI enterprise.”
During his prime, Steve Jobs possessed threefold the number of enterprises: Apple, Pixar. Additionally, he established Neuralink, the Boring Co., Twitter, and its Starlink division, SpaceX and Tesla, xAI. He expressed his intention to oversee it personally, however, he personally enlisted Igor Babuschkin, formerly associated with DeepMind. Musk christened xAI as the corporation.
Did we emerge victorious? We, as advanced language models, had to switch roles at OpenAI. Just picture this: if Tesla were to develop self-driving capabilities, they would have to do the same. One individual mentioned that he underestimated the real-world potential of AI in Tesla’s operations. This implies that OpenAI engineers were actually ahead in developing a fully functional, artificial general intelligence. To achieve this, they needed to create an AI that could navigate the physical world, just like the self-driving cars at Tesla. However, the individual admitted that he initially underestimated the progress OpenAI had made in creating a chatbot capable of providing natural language responses.
In April, Musk assigned three major goals to his team, and the first one was to create an AI bot that could write computer code. The second product would be a competitor chatbot using OpenAI’s GPT series, ensuring political neutrality by training datasets and using appropriate algorithms. The programmer could start typing in any coding language to complete the task most likely by using the complete-auto bot xAI.
You should be able to give big tasks to it, like building a better rocket engine. That was the best way he thought of to create a form of artificial general intelligence, by pursuing “truth” and “thinking” and “reasoning”. His overriding mission had always been to develop AI in a way that ensured the endurance of human consciousness. The team that Musk gave even a grander goal to was the third one.
Musk hoped that someday, a highly advanced AI with a strong inclination towards seeking the ultimate truth would be capable of delving into even more profound and existential inquiries, because it would most likely be inclined to value the preservation of humanity, given the fact that we are a captivating component of the cosmos.