All attempts to regulate the field of artificial intelligence were at best amateurish, states the Knesset's research division. New regulations must be built and an adequate budget allocated to them
Continued reporting on the recommendations of the Knesset's research division. For the first part
Government Resolution 212 of August 2021 tasked the Minister of Innovation, Science and Technology with leading the government's policy in the field of artificial intelligence in the areas of regulation, information and data policy, ethics, civil international cooperation and assimilation in the civil public sector, and more. In order to realize these goals, it was decided to establish an inter-ministerial team led by the Director General of the Ministry of Innovation, Science and Technology that will deal with the issue of artificial intelligence and submit recommendations in the field. The decision also states that the TLM (National Science Infrastructures) forum outline for artificial intelligence and data science must be approved in accordance with the recommendations of the committee appointed by the forum (see below). The recommendations of the TLM outline include: "Development of research human capital in the core areas of artificial intelligence technology ; The establishment of a supranational calculation center for the use of the broad public sector, industry and academia; promoting natural language processing abilities in Hebrew and Arabic; Creating a regulatory environment enables the development of these areas, and more."
In July 2022 it was announced that Israel formulated a "national plan for artificial intelligence" under the leadership of the Ministry of Innovation, Science and Technology and the Innovation Authority operating within the Ministry. The aforementioned national plan was formulated following the work of the Artificial Intelligence and Data Science Committee, which submitted its conclusions to the chairman of the Telam Forum in December 2020. Even before the work of the aforementioned committee, work was also carried out by other experts, including a report on "Ethics and Regulation of Artificial Intelligence" which was written as part of the National Initiative for Intelligent and Safe Systems, and the report "The National Initiative for Intelligent and Safe Systems to Strengthen National Security and Scientific-Technological Resilience: A National Strategy for Israel." The project team began working on the issue in 2018, but the report was only published in 2020. It should be noted that in the outline proposed as part of the aforementioned national project, it was recommended to budget the project at 10 billion NIS for 5 years, and to establish a dedicated national administration to manage it in the Prime Minister's Office. On the other hand, in the report of the Artificial Intelligence and Data Science Committee, the amount of the required allocation was significantly lower - about NIS 5.2 billion, and it was proposed to base it on existing entities.
It should be noted that although as mentioned it was announced that a "national plan for artificial intelligence" was drawn up based on the recommendations of the TLM Forum, we did not find a full publication of the outline of the plan and it is not clear what the total budget for the plan is. According to Dr. Ziv Katzir, director of the TLM program for artificial intelligence, the total investment in the outline is NIS 2024 billion in two installments (in the budgets approved so far, including budgeting) 2023-XNUMX
The Israeli national plan
To the dismay of those who took part in the planning of the national project for intelligent systems, the limited scope of the resulting national plan harms Israel's ability to position itself as a leading, influential player with market power in the field of artificial intelligence - similar to Israel's power in the cyber field, which stems from the early detection and concentration of effort in the cyber field. The Prime Minister, MK Binyamin Netanyahu addressed the issue on June 5, 2023 and noted that in the coming days he will convene think tanks to discuss the national policy of the State of Israel on the subject of artificial intelligence, both in the civil and security fields. This is with the aim of making Israel a national power in the field of artificial intelligence, as it is in the field of cyber.
The formulation of the resulting national policy was done, among other things, against the background of the description appearing in the report of the Artificial Intelligence and Data Science Committee, according to which there is a considerable gap in Israel between the advanced research and development (hereafter: R&D) in the fields of artificial intelligence and a lack of government strategy in the field, which is also reflected in the lack of appropriate computing infrastructures. The authors of the report came to the conclusion that "there is a critical need to initiate a national plan that will outline a systemic solution to promote R&D in the field of intelligence
artificial intelligence and data science. The State of Israel has the potential and the opportunity to be a leading country in the field. The existence of a national program in the field of artificial intelligence and data science is essential to the resilience of the State of Israel and will contribute to the deepening of academic research, the expansion of the industrial base, the creation of significant economic opportunities and the dramatic advancement of Israeli security capabilities."
In October 2022, the Minister of Innovation, Science and Technology at the time published a policy document for public comments under the title "Regulatory Policy and Ethics in the Field of Artificial Intelligence." According to the document "there is no place at this time to promote regulation through horizontal legislation that is unique to the field of artificial intelligence. It should be noted that it is possible that this policy will change as the field of artificial intelligence develops and becomes established and its meanings become clearer."
Among the ethical principles detailed in the document: artificial intelligence will be used to promote growth, sustainable development and Israeli leadership in the field of innovation; The person at the center: respecting fundamental rights and public interests; equality and prevention of unfair discrimination; transparency and explanatory; Reliability, durability, security and safety, and liability.
In a position paper of the Israel Democracy Institute on the subject, it is claimed, among other things, that: 1) the approach in the government policy document according to which a soft and ethical regulation alone is sufficient is not consistent with developments in the world and will lead to the opening of an unnecessary gap between Israel and the developed countries; 2) The document ignores the issue of artificial intelligence uses in the public sector and focuses on regulating the private market; 3) The reference to the regulation of artificial intelligence in itself, ignoring the regulation of tangential technologies, such as privacy and cyber and the failure to regulate them - is problematic.
Government Resolution 173 of February 2023 under the title "Strengthening Israel's technological leadership" approves the national program for artificial intelligence including: acceleration of basic and applied research in the field; Creating a leap forward
for the Israeli industry in the development of infrastructures for the field and implementation of artificial intelligence applications in the civil public sector for its optimization. According to the decision: the Ministry of Finance will allocate a budget for the authorization to commit that will not exceed 500 million NIS, and the bodies that are members of the TLM Forum will prepare the outline budget intended for allocation by them and it will be realized in the years 2023 to 2026. As mentioned, according to Dr. Ziv Katzir, Director TLM program for artificial intelligence, the total investment in the outline is NIS XNUMX billion
beats It should be clarified that the document did not review the government's actual activity in the fields of artificial intelligence, but among other things, support of approximately NIS 5.5 million was approved for the creation and accessibility of databases in Hebrew and Arabic that are supposed to serve as infrastructure for language models; Calls were published calling for the development of software tools in the fields of supercomputing and artificial intelligence and for the establishment of an R&D laboratory for informativeness, privacy preservation and distributed learning, an inter-ministerial team was established to examine the use of artificial intelligence in the financial sector; And a memorandum was signed for cooperation between the Knesset and the Innovation Authority in the fields of natural language processing, based on the accessibility of the Knesset's databases.
4. Artificial intelligence: preliminary information on regulatory initiatives in the world
While as stated above, the last official government position published on the subject supported soft regulation and not a comprehensive legislative initiative on the subject, in Europe there is a comprehensive draft law on artificial intelligence. According to a publication of the European Parliament
The main points of the draft legislation include a distinction between different levels of risk posed by artificial intelligence systems. For example, systems with a very high level of risk to humans, including: systems that use manipulative or subliminal messages, exploit humans or are used for "social ranking" (according to behaviors, socio-economic status, etc.) will be banned from use.
Other prohibited uses proposed in the draft legislation also include: real-time remote biometric identification in public places; Retrospective remote biometric identification, except for law enforcement authorities and in circumstances of serious crimes and subject to legal approval; Biometric sorting based on sensitive characteristics (such as gender, ethnicity, citizenship, religion or political position) "Predictive policing systems" (Predictive "emotion detection" systems for law enforcement, border management, in the workplace and educational institutions, etc.
According to the draft, systems defined as having a high risk include those that harm the health, safety or fundamental rights of humans or the environment; artificial intelligence systems that influence voters in political campaigns; and "systems".
Recommendation" used by social networks with 45 million users or more.
Also according to the draft, suppliers of Models Foundation models in the field of artificial intelligence will have to commit to protecting fundamental rights, health, security, environmental protection, democracy and the rule of law; assess and mitigate risks; Meet various requirements and even register the models in a European database. Basic creative models like GPT will be required with additional transparency requirements including: a duty to disclose that the content was created by artificial intelligence; Designing the model so that it does not consume illegal content and does not publish summaries of copyrighted information or data used to train the model. However, to encourage innovation, it is proposed to exclude from the law research activities and artificial intelligence components that are accessed under an open source license. The law is supposed to encourage controlled experimental environments and "regulatory sandboxes" that will be used for preliminary examination before their deployment.
Unlike the inclusive and active position that is evident in the direction of the regulation of the European Union, so far, a similar framework or concept has not yet been formulated in the regulation in the United States (in this aspect, it is possible that the situation is similar to the different approaches to the protection of privacy between Europe and the United States,) although discussions are also taking place in the United States On the subject of the need for regulation on artificial intelligence.
In February 2019, the White House issued a presidential order titled "Preserving America's Leadership in Artificial Intelligence." The decree establishes five principles: 1) to lead technological breakthroughs in artificial intelligence in all sectors: federal government, industry and academia in order to promote scientific discoveries, economic competitiveness and national security; 2) lead the development of standards and remove barriers to the testing and implementation of artificial intelligence technologies; 3) to train the employees for skill in the development and implementation of artificial intelligence technologies in order to train them for the present and future of employment; 4) foster public trust and confidence in artificial intelligence technologies, and protect civil liberties, privacy and American values; 5) To promote an international environment that supports American research and innovation in the field of artificial intelligence.
Further to the aforementioned presidential order, the President's Office of Management and Budget (OMB) published in November 2020 a memorandum of guidance for the regulation of artificial intelligence applications to the heads of agencies. Among other things, the memorandum states that the agencies must avoid regulation or other actions that would endanger the innovation and growth of the field of artificial intelligence and avoid taking a cautious approach that would set artificial intelligence systems a high and impossible standard that would prevent society from taking advantage of its advantages, and that could harm the United States' position as a leader in the field of artificial intelligence. The memorandum refers to a series of guiding principles for the design of regulation including: public trust in artificial intelligence systems; public participation, scientific integrity and quality of information; risk assessment and management; benefits and costs; flexibility; fairness and non-discrimination; Full disclosure and transparency, safety and security and cooperation between agencies. According to a review by the Brookings Institution on the subject, the approach in the memorandum reflects a policy of reducing regulation on the one hand, and the authorization of the federal authorities and agencies to act in their areas of responsibility with regulatory and non-regulatory tools in accordance with a risk-based approach. Although the memorandum required the relevant government agencies to establish a plan for the regulation of artificial intelligence, only a small fraction of the 41 relevant agencies have done so so far.
The White House intervenes
In October 2022, the White House published a plan for an "artificial intelligence bill of rights" based on five principles: safe and effective systems; protection against algorithmic discrimination; information protection; warning and explanation, and human alternatives. The Brookings Institute's review shows that The aforementioned program focuses on reducing the risks to the economy and human rights due to artificial intelligence systems - unlike the focus of previous government programs on opportunities for society
and to the economy. Also, at this stage the plan is only a recommendation and there are differences in the extent of its implementation in different federal agencies. It should be noted that in January 2023 the National Institute of Standards and Technology (NIST) published a framework document for assessing risks in artificial intelligence.
Sam Altman's testimony in Congress
In May 2023, the CEO of the artificial intelligence company OpenAI Sam Altman testified before the Senate Subcommittee on Privacy, Technology and Law. In his testimony, Altman stated that, in his opinion, regulation of artificial intelligence is essential to stimulate the safety of artificial intelligence systems alongside the protection of people's ability to use them and benefit from them. To ensure that safety is maintained, especially in companies that develop strong artificial intelligence models, it is necessary to make sure that they meet appropriate safety requirements, including internal and external tests before publication. He called on the United States government to consider a combination of licensing or registration of artificial intelligence models above a certain performance threshold, alongside incentives for entities that meet said requirements.
However, due to the dynamism of the field of artificial intelligence, a regulatory regime flexible enough to adapt to technological developments is required and to examine a procedure that integrates all stakeholders to develop and frequently update safety standards, assessment conditions, disclosure rules, validation processes. In addition, according to him, policy makers should examine how to implement licensing regulation on a global level to ensure international cooperation on safety in artificial intelligence and even examine the establishment of an intergovernmental organization for supervision and setting standards. It should be noted that the subject of artificial intelligence is currently being discussed in the American Congress in various contexts and in various committees, and in the meantime discussions are held on the need for regulation in this field. For example, in the hearing of the Senate Homeland Security Committee, the benefits and risks of the use of these technologies by government agencies and the need for appropriate training for their employees were discussed.
More of the topic in Hayadan: