CHATBOTS FAIL TO PASS THE TTT PUB TEST - the TTT CHALLENGE

NLP and NLU forms of Artificial Intelligence (AI ) have a fatal flaw that will fail to pass the pub TTT test of Truth, Trust and Transparency.

An independent, immediate and intelligent national public opinion survey, conducted by trusted,?technically competent humans, totally committed to ethical standards and levels of accountability, matters if we are to gain the benefits of Alternating Consciousness (AC) as a counterbalance to the AI revolution

The emergence of Chat GPT 3.5 represents a significant shift from search algorithms to interactive knowledge expert systems that represent collective consciousness and virtual assistance to students, professionals, and policymakers. Language models have the potential to greatly enhance the way we process and understand information

At the same time, they raise several issues that will enhance the significance of Roy Morgan Single Source stratified samples of changes that promote agency, self-esteem, and the sense of mattering. As language models become more advanced and able to generate increasingly human-like text, there is a risk that they could be used to spread misinformation or to impersonate real individuals online.

Their models are trained on large amounts of data from the internet, which can introduce biases and inaccuracies. This can be particularly problematic when it comes to sensitive topics such as race, gender, or politics that are the essence of both public opinion and market preferences. As these models continue to evolve, it will be important to ensure that they are used in ways that promote accuracy, fairness, and transparency.

There are ways to incorporate large integrated information systems such as national opinion polls, census upgrades, and sales databases to incorporate access to changes in the decision context. One approach is to use these data sources such as Roy Morgan Research to periodically retrain the model, which would allow it to incorporate new information and adapt to changes in the decision context

However, there are many barriers to their unreliable, unaccountable, and regressive sources of informed changes and choices including data privacy, explainability, robustness, fairness and bias, auditability, and avoidance of the issues raised by the robo-debt fiasco.

?While Chat-GPT can generate human-like text, answer questions, and carry out tasks that involve understanding and utilizing natural language, it is limited by the data and algorithms used to analyse anything that happened after mid-2021. Accordingly, it is likely to make mistakes or have gaps in current knowledge when it comes to changing levels of public opinion such as levels of consumer confidence, trust in government, satisfaction with service delivery and appreciation of rare or niche topics, or when dealing with more complex or abstract language.

Another limitation is the problem of interpretability, even though a language model theoretically provides accurate answers. it can give different answers to the same question. it is hard to understand how it arrives at these different answers, makes up its own facts to fill in gaps in its contemporary knowledge and has no common-sense appreciation of the truth or otherwise of its responses.

By all means let’s accept the genius that is emerging as the genie beyond the web, but retain our commitment to the Triple T standards.