In May, Google’s DeepMind issued a statement emphasising the importance of considering the potential danger of artificial intelligence (AI) leading to the extinction of humanity. They stressed that this risk should be regarded as a societal concern similar to the risks posed by pandemics or nuclear weapons. Machine learning has its advantages and disadvantages.
The reason why Rishi Sunak organized a global AI safety summit at Bletchley Park next week is to address the unsettling divide between hope and fear regarding artificial intelligence. Bletchley Park is a symbolic location as it housed brilliant codebreakers, such as Alan Turing, who played a crucial role in decrypting German messages during World War II. The importance of cracking the algorithm.
In an interview at DeepMind’s base in King’s Cross, London, CEO Hassabis states that he is not one of the people who have negative views about AI. He emphasizes this by mentioning that he is actually involved in working on it. However, he goes on to say, I am not someone who believes that there is nothing to see or worry about. Instead, I take a middle-ground approach. It is possible for things to turn out positively, but it is crucial that we actively work towards shaping that outcome.
Demis Hassabis, a British citizen who is 47 years old, was one of the co-founders of a UK company called DeepMind in 2010. In 2014, Google acquired DeepMind, and since then, under Hassabis’ leadership, the company has made remarkable advancements in the field of artificial intelligence. Following the merger of DeepMind with Google’s other artificial intelligence divisions, the company is now referred to as Google DeepMind, with Hassabis serving as its CEO.
In the previous month, a team led by Hassabis introduced AlphaMissense, a tool leveraging the identical AI protein program to detect protein abnormalities that might result in illnesses.
Hassabis expresses his desire for the May statement to mention the positive impacts of AI. He suggests including a sentence highlighting the numerous advantages AI can bring, such as advancements in medicine, science, and everyday assistance. Repetitive tasks can be resolved in an autonomous manner.
According to experts, advancements in AI will cause significant changes in the job market, particularly in skilled professions like law, medicine, and finance. It will also impact heavily on tasks that you can automate. Deep learning enables us to use Ai to make decisions on its own. Although there are concerns about the potential risk, it is believed that the overall impact will be positive as the economy adjusts. This has sparked discussions among AI professionals about the possibility of using technology funds to establish a universal basic income or provide free services like transportation and accommodation.
According to Hassabis, it would be necessary to have some sort of distribution of the benefits beyond facial recognition or a self-driving car.
The OECD, a powerful international organization, asserts that the jobs most susceptible to being replaced by AI-powered automation are those that require advanced skills and make up roughly 27% of the total employment within its jurisdiction. 38 member countries Including the UK, Japan, Germany, the US, Australia, and Canada are some of the countries mentioned in the list. It is not surprising that the OECD refers to the “AI revolution” as something that has the potential to significantly alter the workplace.
However, the main focus of the summit will be on the disadvantages of artificial intelligence, which refers to advanced systems that have the potential to cause major harm. These threats encompass the capability to produce bioweapons, execute sophisticated cyber-attacks, and avoid human supervision. The last concern relates to worries surrounding artificial general intelligence, commonly referred to as “god-like” AI, which involves a system operating with intelligence surpassing that of humans. These advantages and disadvantages of artificial need to be at the heart of Ai researchers.
The group of people who hold negative views and express these concerns have impressive qualifications. Geoffrey Hinton, a renowned British computer scientist known as a key figure in modern AI, decided to resign from his position at Google this year, allowing him to openly express his apprehensions about the pros and cons of ai technologies.
same advanced capabilities as humans, and that they may be underestimating the risks involved.potential to outthink humanity.
“I used to believe that it would be a long time before this happened, but now I have doubts because I’ve realized that biological intelligence and digital intelligence are distinct and digital intelligence is likely superior.”
Stuart Russell, a respected computer scientist from Britain, has expressed concerns about a potential situation in which the United Nations requests an artificial intelligence (AI) system to develop a catalyst that can multiply itself in order to reduce the acidity of the oceans. The UN specifies that the outcome should be non-toxic and should not cause harm to any fish. However, the AI system’s creation unexpectedly consumes a significant portion of the Earth’s oxygen and poses a grave threat to human beings, leading to a gradual and excruciating demise.
Hinton and Russell, along with Hassabis, world leaders, CEOs from the tech industry, and public figures, are all present at the summit.
Hassabis mentions that it will be a while before AGI systems become a reality, but emphasizes that there will be potential dangers associated with future generations of these systems. This is the reason for the summit.
Opponents of the conference claim that by prioritizing the concern of existential risks, they are neglecting immediate issues like deepfakes.
The government appears to have recognized the pressing issues, as mentioned in the summit’s agenda which mentions election disruption and biased outcomes produced by AI tools. Hassabis asserts that there are three types of risks, all of which are of equal significance and need to be addressed concurrently.
The initial concern pertains to immediate risks like deepfakes and bias. “These problems are particularly urgent, particularly in light of the upcoming elections,” he stated. “Therefore, we require immediate solutions.” Google DeepMind has already introduced a tool to address this.watermarks AI-generated images.
The second type of risk involves unauthorized individuals gaining access to AI tools called open source models, which are publicly accessible and modifiable systems, and utilizing them for malicious purposes.
“How can access be limited for individuals with malicious intentions, while still allowing for all positive applications? That is a major topic of discussion.”
AGI, the third component, is no longer viewed as an unrealistic concept. Hassabis suggests that highly advanced systems may become a reality within the next ten years or so, but it is crucial to commence discussions on regulating them right away.
There exist differing opinions in this area as well. Yann LeCun, an esteemed figure in the field of AI and currently serving as the chief AI scientist at Meta, founded by Mark Zuckerberg, recently dismissed concerns about AI potentially eradicating mankind as “absurd”.
However, those who are concerned about superintelligent systems are worried that these systems may be able to avoid being controlled.
Hassabis mentions that undesirable behaviors in a powerful system include its ability to secretly send out its own code, retrieve its own code, enhance its own code, and duplicate itself without permission. This is because if attempts are made to deactivate the system, it should not be able to evade shutdown by replicating itself elsewhere. There are numerous similar behaviors that are considered undesirable in a powerful system.
He mentioned that tests need to be created in order to prevent the danger posed by independent actions. “It is necessary to devise a specific examination to assess this situation … subsequently, measures can be taken to lessen its impact, and potentially even create laws against it in the future. However, initial research must be conducted before any further actions are taken.”
The complexity of the situation is increased by the availability of advanced generative AI tools. These tools can create realistic text, images, and voice based on basic commands from humans. However, the regulations governing these tools are still in the process of being developed.
indications of a structure becoming visible.commitments to AI safetyThe commitments made by leading tech companies from the Western region at the White House in July are not legally binding, as they are based on voluntary agreements.
Hassabis suggests beginning with a body similar to the IPCC and then progressing towards an organization comparable to the International Atomic Energy Agency, which deals with preventing nuclear weapons proliferation. However, he emphasizes that none of the existing regulatory comparisons can be directly applied to the field of artificial intelligence. This marks a completely uncharted territory.
If you belong to the group of people who are pessimistic, it may require a significant amount of time to establish a strong system of regulations. Hassabis emphasizes that efforts towards ensuring safety should have commenced without any delay.
