Despite longstanding fears about AI, NTU researchers managed to compromise AI chatbots by “jailbreaking” them to produce content which breached their developers’ guidelines, exacerbating the ambivalence many have felt about AI.
In an interview with Lianhe Zaobao, Hong Yin Yin, CEO of Elite Asia, a local translation agency combining human and machine translation services, shared that in the past, it was extremely challenging to translate over 100 pages of legal documents in an evening, but AI has however increased translation efficiency greatly.
“The ‘jailbreaking’ problem identified by the NTU research institute might cause AI chatbots to produce false, if not even harmful information that would be to the detrimental to companies. Companies using free-of-charge AI chatbot services might unintentionally expose confidential corporate information to external parties, hence creating conditions for ‘jailbreaking’.” Fong believes that to for this problem to be avoided, having private servers and human safeguards are of utmost importance.
Li Weina (35 y.o.), a proofreader with 8-years’ experience, shared with Lianhe Zaobao that although machine translation offers advantages which include completeness and consistency in use of terms when compared against human translations, she is also worried that AI may be used to fabricate falsehoods or even create weapons.
“At the Global Media Congress in the UAE, some exhibitors have demonstrated how a person’s photo and voice recording is all that is needed for him to report fake news in different languages.”
Chen Weijie (39 y.o.) is a Singapore teacher with 16 years’ teaching experience who graduated with a Computer Engineering degree from the National University of Singapore. An adept educator who deftly integrates mathematical models and technology into teaching, he had heard about the NTU “jailbreaking” but was somewhat disturbed by NTU’s findings.
“NTU’s demonstration fully showcases AI’s exceptional self-learning capabilities, yet also brings to the fore the dangers of AI being manipulated to breach ethical and operating guidelines,” Chen said. He adds that we must remain vigilant towards AI-enabled communities, be ethically forward-looking and solve problems collaboratively to ensure that AI improvements do not weaken social and ethical frameworks.
Former Google Executive Warns of the Risks AI Poses
When interviewed, Laurence Liew, Director of AI Innovation at AI Singapore (AISG), pointed out that Geoffrey Hinton, a former Google executive, along with other scientists, had well-founded grounds for cautioning of the risks AI poses. If AI’s disruptive capabilities are used incorrectly, it may be reduced to a tool for spreading fake news, launching cyber-attacks, and even used to generate bigoted content en masse.
“We must remember that AI like all other powerful technologies, and what matters is whether these technologies are used correctly.”
Liew shared with reporters that AI Singapore is aggressively developing AI smart products and solutions and will use these projects to kickstart the AI Apprenticeship Programme (AIAP) targeted at Singaporeans. All these projects are rigorously evaluated so that apprentices are well aware of the ethical and safety issues related to AI.
Published: January 03, 2024, 09:03 PM SGT