Deep Fakes, AI and the New Electronic Struggle

THE PRESIDENT’S youngest son Tumelo Ramaphosa recently appeared on national television, punting blockchain technology, artificial intelligence (AI) and a digital financial future. Some of his previous projects include turning wildlife into digital tokens via a crypocurrency investment scheme for stud farmers called StudEX, and a swathe of more ephemeral ways to fleece (some might say leverage) the startup space in San Francisco.

Apparently drones to track SA wildlife conservation efforts are being funded in part by auctioning off Dad’s bulls via cryptocoins. Aside from the unfair advantage that comes with being the president’s son, one can think of far better ways of spending one’s pocket money than reducing animals to mere fractions.

As Elon Musk stated during his encounter with Jack Ma, ‘don’t assume that artificial intelligence research is being conducted by intelligent people’. Indeed there are many critical and exceedingly dystopian concerns about the emerging paradigm that looks set to surpass humanities ability to comprehend the impacts of AI.

A recent research paper published by Yale fellow Michael Kwet paints a rather bleak picture of how smart CCTV networks are driving an AI apartheid. 

In the process video analytics are reinforcing racial and class divisions,  creating a world in which the poor are lo and behold, excluded by the rich. The latest round of criticism has an eerie similarity to my complaint made to the US press back in the 1980s. In a letter published under a pseudonym and carried prominantly by cyberpunk magazine Mondo 2000 I outlined the manner in which the apartheid regime had weaponized the banking sector, deploying ATMs as a convenient means of entrapping activists.

The complaint predated the later unsuccessful IBM case brought alongside a suite of apartheid litigation against Ford and other US companies, which unfortunately never made it out of the starting blocks, thanks to overly broad generalisations in the founding papers, lack of public interest here and abroad, and a US second circuit decision striking down the Alien Torts Act.

Suffice to say, that IBM were most certainly responsible for the technology behind the Dompas and thus apartheid race classification technology.

The mind shudders to think what would have happened if the apartheid state had access to AI technology, although somehow I like to think that the anti-apartheid movement would have probably hooked onto blockchain and crypto in the same way that Rhodes Must Fall/ Fees Must Fall took to social media.

If the thought of racist rednecks weaponising AI is a little disturbing, a recent news article warned that a group called OpenAI had ‘declined to release research publicly for fear of misuse.’ Apparently “the creators of a revolutionary AI system that can write news stories and works of fiction – dubbed “deepfakes for text” – have taken the unusual step of not releasing their research publicly, for fear of potential misuse.

The ability to fake and control news in the process driving public influence is not a new concern, as movies such as Videodrome (1983) and Network (1976) have already pointed out.

What is new, are privacy concerns such as mine, about the potential of AI to unlock passwords, defeat cryptography, and reassemble data in new and innovative ways. On the flipside, AI will improve our understanding of past civilisations, forgotten languages and art. Like any tool it may be used for good and bad.

A Japanese research team using AI recently uncovered some 2000 new Nasca lines, previously invisible images in the Peruvian desert.

Back home, this jump in processing power, represents an incredible opportunity to recover ancient memory lost to pre-colonialism. Settlements such as Mapungubwe and Great Zimbabwe have a lot to offer. Deciphering and protecting texts surrounding the university of Timbuktu, will undoubtedly grow in leaps and bounds, but not if AI is simple code for venture capital and used tech salesmen,  while our nation’s research institutions are quietly stripped of intellectual capital.

Again, AI for all its scifi brohaugh is really a misnomer, the correct and better phrase is ‘machine learning’. Its a synonym for AI, not a sub-set as in ‘reinforcement learning’.

In particular, the terrain of intercultural communication could take off in significant ways, if our country were to set national goals, for instance providing each and every citizen with the tools to communicate across the linguistic divisions which have traditionally acted as hurdles to our understanding of each other.

As an individual afflicted by the presence of several African languages each competing for his or her attention within my own household, I can only hope that instead of auctioning off bulls via blockchain like our president’s son, our nations youths, instead present us with with a workable plan to grant each and every South African the benefit of instantaneous machine translation — a fact of life still missing from the Southern African region, but surely one that will become a boon in the future?

UPDATE: Ramaphosa’s son slammed for cancelling conference without notice

Ramaphosa Junior’s blockchain belly-flop