Artificial Intelligence CompetitionIntroductionArtificial intelligence (AI) is affecting many sectors of the economy in a time of rapid evolution. There is no doubt that governments and politicians have a prominent role in understanding and addressing AI and its potential impact. To manage potential competitive harms and to promote competition in AI markets, the Competition Bureau (the "Bureau") constantly researches to develop a better understanding of AI, how it may affect the competition, and what needs to be done to prepare for it. On this issue, the Bureau participates in intergovernmental cooperation by interacting via the Canadian Digital Regulators Forum with the Office of the Privacy Commissioner (OPC) and the Canadian Radio-Television and Telecommunications Commission (CRTC) to discuss the implications of emerging technologies, such as artificial intelligence (AI), as they relate to our respective mandates. Artificial Intelligence and CompetitionAI and mergers and monopolistic practicesThe Competition Bureau investigates mergers that have the potential to reduce or eliminate competition significantly and takes appropriate action. Specific features of AI marketplaces may influence the emergence and development of market power and concentration. Barriers to entryThe ability to get data and compute inputs is a prerequisite for involvement in AI development marketplaces. For new businesses looking to enter the AI development industries, acquiring the necessary data or computing power may be a barrier to entry. Public data that may be utilized for AI applications (like training an AI model) is quite common and now makes up most of the data used to create contemporary, well-known AI technology. This suggests that the majority of the data inputs needed for new competitors to participate in the market could be easily accessible. Proprietary data, however, is already crucial to several phases of the development of generative AI and might become more so in the years to come for the development of AI more generally. If access to private data becomes a prerequisite for involvement in AI marketplaces, this might put some barriers in the way of successful competition for newcomers. While the requirement for vast amounts of data, especially private data, to compete successfully may make data access a barrier to entry, advancements in AI techniques and technology may reduce the demand for data in AI development. For instance, enhanced model architectures or novel training techniques may result in efficiencies that use fewer data resources to accomplish the same level of performance. Businesses may employ synthetic data more frequently to advance AI. Using synthetic data might be less expensive and easier to get than gathering public data or purchasing private data. However, there are disadvantages associated with using synthetic data, including the possibility of bias propagation and data contamination, which some research has indicated might lead to "model collapse." Although there is a growing market for data providers for AI applications, it is still being determined how these markets will evolve in terms of competition and market strength, and hence the price of access that downstream AI development businesses must bear. Big IT companies might have access to significant amounts of confidential data that can be used for AI development because of their involvement in other areas. Businesses that already have massive data storage may benefit from this in terms of competitiveness. It has been discovered that the marketplaces for AI chips and cloud computing are highly concentrated, with a small number of very large technological companies dominating each sector. Due to the substantial initial expenses of computing inputs, new businesses wishing to enter the AI development or infrastructure industry may need help getting started. Startups would probably have to obtain the computational inputs they need from a major technology supplier, which might also include competitors in the AI space. The current shortage of GPUs used for AI applications may provide further challenges for new participants in the AI development marketplace. Product DifferentiationBusinesses can use AI to set their goods apart from those of competitors, for instance, by adding AI features to already-existing items or using AI to create new, improved products. In marketplaces where businesses offer unique items, the possibility of coordinated behavior is reduced, which might be advantageous to competition. Companies can gain market power through product differentiation. Compared to marketplaces where many companies provide similar items, distinct markets may see less intense price competition. However, the market dominance obtained by product differentiation could not last when new inventions joined the market and competed. Economies of scope and scaleSignificant initial expenses lead to economies of scale in the marketplaces for AI development. Because they already have internal resources for computation and human experience, large technology companies may be able to achieve economies of scale in the AI industry. However, it could be tough for new market entrants or competitive players to enter the market, as they have different advantages. The vast range of AI applications across different sectors and use cases may allow businesses in AI marketplaces to realize large economies of scope. Given that foundation models are known for their flexibility in handling a wide range of downstream tasks, this effect could be especially pronounced in foundation model marketplaces. As a result, markets may become more concentrated among businesses that can engage in a variety of foundation model applications. Network effectsNetwork effects are frequently used to describe digital marketplaces. AI marketplaces could offer environments where data network effects or indirect effects might occur. With so many downstream applications, markets for foundation models or generative AI might result in indirect network effects, where the more downstream applications these technologies are utilized for or incorporated into, the more valuable they become. According to a recent study, there is a correlation between artificial intelligence (AI) and data network effects. This means that as an AI system gathers more user data, these technologies become more successful and draw in additional users who generate data. Predatory, exclusionary, and discriminatory conductWhen a business takes temporary losses to get rid of a rival and increase its market share, it is acting predatorily. AI might help businesses find greater possibilities for predatory behavior and put predatory plans into action. AI may, for instance, be better able to recognize consumers who are likely to transfer providers and implement tailored pricing, focusing on offering "at-risk" consumers lower-than-market prices while maintaining the same level of service for others. By minimizing losses on below-margin consumers, this tactic would reduce losses to the company using predatory behavior, which would raise incentives to participate in such behavior. When one company attempts to punish another, it is acting in a disciplinary manner. For example, in AI marketplaces, lower prices might be used to punish competitors instead of simply eliminating them. Exclusionary activity is the act of a company trying to keep another company out of the market. The significance of vertical linkages in AI marketplaces could create a setting where some forms of discriminatory behavior could emerge. Vertically integrated companies have the potential to compete in the same downstream markets as the companies they provide to while simultaneously providing critical input for engagement in AI marketplaces, such as data, computing, and AI models. If these companies were to act in a way that would drive out their downstream competitors from the market, such as by refusing to supply or by reducing the difference between the prices they charge for inputs they supply and the prices downstream businesses charge, it would be problematic for competition and create barriers to entry for downstream businesses. By giving their items an advantage over competitors, AI might be used by platforms that participate in self-preferencing. It could also make it harder for users to identify sponsored suggestions. With vertically integrated companies that compete in AI development markets and provide significant computational inputs like cloud computing services, the AI markets may provide a platform for discriminatory behavior to emerge. Large technological companies that compete in several digital marketplaces are a defining feature of AI markets. These businesses could use bundling and tying, which are popular tactics used in a variety of sectors and are often linked to significant cost savings. Certain packing and tying techniques, however, can be considered an abuse of market power. For instance, companies that are already dominant in one market may use bundling and tying tactics to become even more dominant in other sectors where they compete. AI in and of itself can be used to target tying and bundling methods by focusing on consumers who are likely to switch. When artificial intelligence (AI) is used to make decisions, such as setting a company's pricing strategy, there can be particular issues. For instance, AI could decide on its own to engage in predatory, exclusionary, or discriminating behavior in the absence of human guidance. Businesses that rely on AI to make decisions need to make sure that enough human monitoring is in place to spot and address any instances in which AI decision-making may be anti-competitive. Considerations for mergersAs part of its merger evaluation process, the Bureau evaluates whether a combination will likely result in significant price increases or the creation, maintenance, or enhancement of market power. If the merger gives the merged company a dominant position in the market, the merged company may be able to engage in exploitative, exclusionary, or discriminatory behavior. In horizontal mergers, two companies that compete in the same market or markets may combine their market shares or other resources to create a merged company with significant market power. Firms operating at different levels of the supply chain are combined through vertical mergers. As a consequence of a vertical merger, customers or suppliers of a company may become rivals, which may increase the company's position in the upstream or downstream markets. The Bureau would be concerned if a vertical merger may lead to or enhance competitors' capacity and incentive to restrict or completely remove their access to markets or inputs. Conglomerate mergers may provide rise to the possibility or motivation for the merged company to use anti-competitive tying tactics in cases where it competes in several connected markets after the merger. Because these sectors are already highly concentrated, any merger involving a company that provides computational inputs such as AI chips and cloud services may require more examination. Due to the possibility that large, established companies would try to buy up new rivals in order to avoid or reduce competition, mergers in the AI space may need more examination. If a merger allows or promotes coordinated activity between enterprises after the merger, it may effectively prohibit or significantly reduce competition. Companies that participate in accommodating activities that benefit all coordinating companies are said to be engaging in coordinated behavior. Businesses may work together to decide on a range of commercial matters, such as pricing or market distribution. It is possible to create coordination through implicit understandings rather than constantly requiring clear communication. Artificial intelligence (AI) has the potential to support coordinated behavior by gathering data about rival behavior and intervening quickly. ConclusionBy reviewing the AI competition, we have understood the vast areas and fields that AI is capable of implementing in real life. People showcased possible answers to many challenges within different domains, including artificial intelligence enhancement, increasing efficiency in natural language processing, and computer vision. The competition encouraged the integration and advances of current state-of-the-art technology, thus unraveling the revolutionizing power of AI all across industries and the social realm. It also highlighted the necessity of ethical concerns and the need for the proper societal production of AI. In total, the event was proof of the vast improvements in the area of AI research and possible developments in the furthering of the subject, urging for more elaboration regarding this fast-growing and crucial topic. Next TopicAudio Machine Learning |