Rapidly Evolving Conversational AI Shiny or Scary?
With the success of ChatGPT, AI moved forward in two important ways: AI has scaled massively in its recognized use in society and in the size of data and complexity of models. Between November 2022 and January 2023, the ChatGPT issues most discussed on social media were: Content creation (digital marketing, search engine optimization); Education (professional training, increasing productivity); and the potential of Generative AI (approximating human cognition, conducting innovative research, potential use in digital counseling and mental health).
Chat GPT-4 has an unparalleled 1 trillion parameters. It is estimated that running ChatGPT costs OpenAI $3 million per month, but it is available for the masses. ChatGPT's output sounds very confident, but factual accuracy is still limited to 70% to 80%. To reach higher accuracy, it will require even more data and compute power, which are unlikely to be enough. Specifically in software development, ChatGPT can be used for faster code development and quick bug fixes; improved customer support and response systems; and higher developer productivity on AI tools.
According to Gartner current AI does not outperform people in most tasks yet. AI is less advanced and less powerful than most people realize. AI can tackle some narrow tasks far better than humans, but most work involves multiple skills, broad knowledge, reasoning and context switching that only humans can manage.
AI in 2023 provides an opportunity for each of us to become more productive and consistent by expanding the reach of our expertise, creativity and knowledge. ChatGPT’s ability to simulate massive quantities of models for creating scenarios will enable AI to accelerate new insights, models and capabilities. In the future, AI may displace workforce tasks, but it remains to be seen whether AI can replace people entirely.
AI augments human capability and replaces worker tasks. Humans augment AI capability, creating new jobs, while helping AI applications with tasks that AI cannot do on its own. Moreover, unchecked or unquestioned, AI’s presumed “impartiality” can actually make it a bias amplifier. It is very difficult to nigh impossible to reverse-engineer an AI system once it’s been trained, making auditing these systems challenging.
AI augmentation strategy must fit within the expectations of the workforce. AI can help sales associates at retail stores recognize not just who has been identified already as a high-value customer, but also who fits the profile of a high-value customer, but hasn’t been identified yet. Also AI can help associates in high-volume stores make effective product recommendations. AI can put knowledge into context and calculate likelihoods for decisions outcomes that reduce the time to reach a final decision.
The business environment changes constantly and requires more and more flexibility and learning as you go. For this reason, humans remain the best way to create value for customers and citizens. The enterprise will always need workers. AI augmentation supports adaptation to changing conditions by helping people with new competencies and with mitigating new risks. AI augmentation can turn good workers into great workers. AI can even guide workers to change behaviors when the enterprise’s strategy and tactics change.