In recent weeks we’ve seen Microsoft’s ChatGPT rollout of integration with their search tool Bing prove problematic. What was offered to the public turned out to very much be a Beta Version, with many bugs and even the overall approach needing work. As became evident, MS rushed to market to get ahead of Google’s Chatbot AI which was also rushed to market and publicly failed its big Demo unveiling. Perhaps such “we’re going to be first and grab the market share” tactics were bound to fail with such a complex tool such as current Chatbots offer.
MS has since tried to corral their ChatGPT tools into something much more controlled, but the nature of the beast seems to “want to be free”. Control seems to be a high price to pay for something whose benefit is all about AI control over information returned from queries.
Google will eventually get their chatbot AI together, but will face the same questions re control of the AI beast. Generally it seems there’s no escaping the reality that giving “control” to AI is problematic from many different perspectives. Ray Kurzweil says we don’t have to kill the beast because of its potential to go rogue and make choices not based on human interests. He thinks we will always be in partner with the technology, and will always have the last word on how AI is used.
Given the initial rollout problems of ChatGPT et al, it’s harder to see Kurzweil’s position. For AI to really work best, it seems that we need it to surpass human limitations, and when we do that, we cede at least some portion of “control”. The trick is likely to release control of some aspects, but not to the point of losing overall control. How that mix can be derived in a competitive Tech environment, when the regulatory functions of government cannot hope to “catch up”, is a very challenging conundrum.
Given the current tech case before SCOTUS, one is hard pressed to be optimistic that problems will be “headed off”. Rather it appears we are going to be doing a large amount of “off-road” travel on the information superhighway across iffy and in some cases treacherous terrain. But that’s not new. The advent of Nuclear Power technology has posed the same question of control since it was first created. Which remains: can we control the negative results, while obtaining the positive results?
We seem no closer to escaping that conundrum today.