When Open AI introduced ChatGPT, it appeared to be a promising tool, akin to an oracle endowed with vast knowledge from all corners of the internet. Such a resource could have been invaluable in a world grappling with polarization, misinformation, and a widespread erosion of trust in society. Alas, the limitations of this technology quickly became apparent. ChatGPT had a proclivity for generating hallucinated answers, devoid of any objective truth. Furthermore, a flood of chatbots from various companies emerged, each providing different results to the same prompts. With the proliferation of these chatbot models, concerns surrounding biases and harmful content abounded. It seemed the pursuit of truth was becoming more complicated than ever before.
Taming the Chatbot Chaos with Guardrails and Constitutions
As criticisms of ChatGPT’s guardrails and alleged liberal bias surfaced, tech moguls Elon Musk and the team at Anthropic sought different approaches. Musk pledged to create a chatbot that was less restrictive and politically correct, aiming for an alternative that aligned better with different perspectives. Anthropic, on the other hand, implemented a “constitution” for their Claude chatbots, emphasizing values such as helpfulness, harmlessness, and honesty. Incorporating ideas from the U.N. Declaration of Human Rights and diverse perspectives, these constitutions aimed to steer chatbots toward responsible behavior.
Open-Source Models and the Quandary of Guardrails
Although companies like Meta released open-source language models (LLMs) such as LLaMA 2 and Vicuna, effectively enabling free and unrestricted utilization, the issue of controlling potential harms persisted. Recent research unveiled a prompting technique that rendered guardrails ineffective, raising concerns about the accessibility of harmful instructions or fraudulent practices through LLMs. Even with developers’ best efforts to counter such attacks, preventing all forms of misuse remains an unsolved challenge. The consequences of fractured truth and societal fragmentation are dire, echoing the digital chaos pervading social media and news outlets. The rise of chatbots, alongside the swiftly evolving world of digital human representations, further compounds these issues.
From Text-Based to Multimodal: The Rise of Digital Humans
While current chatbots built upon LLMs largely rely on text-based responses, there is a growing trend toward multimodal models capable of generating images, videos, and audio. In the realm of “digital humans,” highly detailed and realistic human models are being developed. These creations interact with real humans in natural and intuitive ways, facilitating applications in customer service, healthcare, and remote education scenarios. Notably, the emergence of digital humans has reached the realm of newscasting, with Kuwait News and China’s People’s Daily already experimenting with AI-powered newscasters.
The Rise of AI-Generated News and Synthetic Faces
Startup company Channel 1 is set to revolutionize video news channels with the help of gen AI and LLMs. Their vision is to create personalized newscasts, catering to individual interests and even filtering news through specific ideological lenses. Although the technology is still developing, Channel 1 believes that, within the next few years, the distinction between AI-generated and human-presented news will virtually disappear. However, the potential danger lies in the fact that synthetic faces are often perceived as more trustworthy than real ones, raising concerns about the exploitation of this technology for nefarious purposes.
The Future of Truth and Trust in a Manipulated World
With society already on edge due to disinformation campaigns, voice-cloning, and deepfakes, the advent of manipulated video news amplifies existing concerns about truth and trust. As technology progresses, the risk of distorted information designed to manipulate opinions becomes more apparent. The pursuit of truth and trust has become increasingly complex, and the fragmented nature of digital communication exacerbates these challenges. In an era where the influence of Walter Cronkite’s evening news seems like a distant memory, the need for reliable, unbiased information becomes all the more vital.
While chatbots and digital humans offer intriguing possibilities, the utilization of these technologies must be approached with caution. As the quest for truth and trust continues to face obstacles, it falls to the developers, companies, and society as a whole to strike a balance between innovation and responsible implementation. Only then can we hope to navigate the complexities of the digital age and shape a future where truth prevails over manipulation.
Leave a Reply