Google will not sell general-purpose facial recognition APIs prior to working through ‘important technology and policy questions’, according to a recent blog post from SVP of Global Affairs Kent Walker discussing the use of AI in the Asia-Pacific region. The technology has the potential to benefit society, helping to locate missing persons, identifying participants in illegal activities or when used in medical diagnostics but is not without risks, the executive says. Input and research are needed as well as cooperation from and including governments, non-profit organizations, universities, and businesses in both the public and private sector to avoid abuse of the AI or unexpected outcomes that are ultimately harmful. Understanding that helped to inform Google’s decision to avoid selling its general-purpose recognition-based AI without due consideration for its consequences.
The company is also taking its efforts to prevent misuse of the technology in the region a step further through a grant it is offering toward the creation of the Asia Pacific AI for Social Good Research Network. The organization will act as a discussion network collective of leading academics and discussion with governments, civil society, and the private sector and a hub for research on AI as it pertains to societal well-being. The goal of that network will be to guide AI developments in the Asia Pacific toward the responsible use of the technology that follows principles and values that align with Google’s.
Background: The newly-announced next step in Google’s AI efforts in the region follow previous bids to use AI in medical examinations and uses such as for the diagnosis of diabetic retinopathy. That’s a condition that can lead to permanent blindness in diabetic patients but isn’t often caught early because of a lack of specialists in the eye care industry. Among those projects taken on by the search giant is a pilot program for an AI addressing that diagnostic task that will soon be launched in Thailand. While that isn’t anywhere near the full extent of Google AI programs tying into fields of medicine for use in diagnostics, it provides a clear example for the company regarding the ways AI can benefit society as a whole. In other regions of the world, the company has cooperated with medical professionals and leaders in research for that industry in bids to address various forms of cancer and cardiac problems, among other health issues.
That’s in addition to its bid to unify healthcare information under an end-to-end AI system specifically designed to assist in diagnostic and other care tasks. However, Google has also been under a great deal of scrutiny lately as a result of several decisions relating to data privacy, bias, and allegations related to other internal practices. Most recently, those compounding problems led to the appearance of company CEO Sundar Pichai before a House Judiciary Committee in a hearing to determine whether it was abusing its position as the world’s preeminent search provider and mobile OS maker. Questioning also brought in concerns about whether the company might be breaching its own standards and values by pushing forward with Project Dragonfly. That’s a browser that opponents say would be complicit in the abuse of human rights in China, where laws and regulation are already prevalent that suppress and control the use of technology. More pertinently, the region is notorious for its use of AI and facial recognition to apprehend criminals or to enforce laws regarding restrictions on citizens.
Impact: Google’s position on not selling API’s related to machine vision for facial recognition will likely be seen as a step in the right direction compared to the controversial actions and decisions it has made over the past several months. But it isn’t a complete solution either since it doesn’t specify what the conditions for opening up sales would need to be. In short, it doesn’t limit or prevent the sale of that technology once the search giant has decided that enough investment into research has been finalized or do anything to ensure partners follow a similar policy. There also doesn’t appear to be any consideration in Google’s promise for special-use scenarios or specialty AI as this only applies to general-purpose AI.