[ad_1]
Public conversation around technology governance in recent times invariably proceeds from an axiom: Regulation and policy are at least a step behind innovation. This has indeed been the case historically, from the printing press to the rise of social media. But it need not be so with AI.
The public launch of Large Language Model (LLM) ChatGPT by OpenAI in November 2022 has disrupted the technology space. 2023 was witness to the democratisation of LLMs — never before has AI this sophisticated, with such human-like interfaces and vast knowledge, been so easily available. ChatGPT also brought into the open the race between the biggest companies in the world including Google, and between China and the US, to develop Artificial General Intelligence (AGI). This year saw a watershed moment in technology, as well as questions around its regulation.
There is a curious dichotomy between the so-called dangers of AI and its purported advantages. The latter are specific: From advanced gene sequencing to tutors for students, from writing code to a competent virtual assistant. The dangers are generally apocalyptic, ranging from science fiction-inspired doomsday scenarios to the effects LLMs and AI will have on creativity, work and the economy, and society as a whole. Two possible reasons account for this. First, those who develop the technologies have a fair idea of its capabilities (although the more advanced AI becomes, in terms of computing power and the data sets it is trained on, the more unpredictable it becomes) but perhaps less so about its impacts. Second, the jobs and ways-of-being threatened by LLMs and possible AGIs differ in a fundamental aspect from earlier disruptions. This time, it is white-collar jobs based on intellectual and cognitive abilities that are under threat. That could explain why there was less panic over mechanisation replacing manufacturing jobs than, say, machines replacing lawyers, doctors, artists and writers.
The panic around AI holds twin dangers. First, it could invite a regulatory response that stifles the potential of a tool that can aid innovation and make knowledge and skills more accessible. Second, it may obfuscate the issues that incrementally end up making AI more harmful than beneficial. The counter to plagiarism needs to evolve. Questions of authorship and copyright are already being litigated in some jurisdictions — for example, with The New York Times case against OpenAI. The use of machine learning software for surveillance, facial recognition and predictive policing has major implications for privacy and human rights. Addressing these concerns requires more than the expertise of the engineer and the technologist. In 2024, society needs to drive technology, rather than the other way around.
[ad_2]
Source link
