On Nov. 30, 2022, OpenAI announced it had trained a language model called ChatGPT and it was available for public testing. In the months that followed, ChatGPT created a buzz in the tech community that rivalled some of the biggest leaps forward in recent memory. Think of technological milestones like the iPhone, the Tesla Model S or the astounding ATLAS humanoid robot developed by Boston Dynamics. ChatGPT is on that scale of impressive.
You have to play with this new chat toy to fully comprehend how remarkable it is. Ask it to prepare an essay on any topic, write a software algorithm that you describe or just have a chat about a subject you’re interested in. Naturally, I asked ChatGPT what it thought about clearcutting old-growth forests. The answer was remarkably balanced and well-informed. It was not really what I wanted to hear.
I attempted several leading questions to see if I could get responses more aligned with my point of view. Surprisingly, ChatGPT provided rational and well-thought-out answers, while reiterating the positions of both sides of the issue. For example, I received the following response when I asked if jobs are more important than ecosystem services:
“It is not necessarily accurate or fair to frame the issue as a choice between jobs and the ecosystem services provided by an old-growth forest. Both of these factors are important considerations, and it is important to find solutions that take both into account.
“Jobs are an important source of income and stability for individuals and communities, and the logging industry can provide significant economic benefits, particularly in rural areas. At the same time, old-growth forests provide a range of important ecosystem services, such as water filtration, carbon sequestration, and soil protection that are vital for the health and well-being of both people and the environment.
“It is possible to find solutions that both support economic development and protect the ecosystem services provided by old-growth forests. For example, this might involve adopting sustainable forestry practices that minimize the impact on the environment, establishing protected areas or conservation reserves, and finding ways to value and compensate the ecosystem services provided by old-growth forests.
“Ultimately, the best approach to managing old-growth forests will depend on the specific context and the values and needs of the stakeholders involved. It is important to find solutions that are sustainable, equitable, and respectful of the values and needs of all stakeholders.”
As our nearly two-hour discussion progressed, I began to learn that I could ask for specific references that ChatGPT used to form its responses. This included research papers on how old-growth forests positively impact our well-being and several books describing the spiritual connection between the Haida Nation and the forests of the Haida Gwaii. However, one time when I asked for a reference, I received the reply:
“I apologize, but I am unable to provide a specific reference for this information as it is drawn from my own general knowledge and understanding of the topic.”
The artificial intelligence model used to create ChatGPT is far superior to the simple question, answer and command algorithms used in already impressive voice-activated systems like Amazon’s Alexa or Apple’s Siri. Just think what Alexa could do once it is integrated with the latest AI capabilities. We’re not far from this being a reality in every household, smartphone and vehicle. Companies like Google and Meta are developing their own AI applications and are undoubtedly ramping up their development efforts now that this new chatbot technology is out of the box.
Let’s pause for a moment and think about how we really feel about a future where we have an electronic device in our living room that we can have an articulate and interesting conversation with on any subject. It won’t replace television, but it might help reduce loneliness for a lot of people. I’m personally looking forward to having a talking computer assistant walk me through a recipe, vehicle maintenance or a home renovation project.
There are also concerns about this technology being used to commit fraud or spread disinformation and conspiracy theories. The ChatGPT model seems to do a very good job of providing balanced views, but any software developer can bolt on a front end that filters portions of the response and provides a very unbiased or misleading opinion.
In 2021, the San Francisco Chronicle ran a truly chilling article about someone who used the Project December website to have very realistic conversations with a persona created to represent his deceased fiancée. The programmer who created Project December used an older version of the GPT3 engine that is behind the ChatGPT model. In the end, the simulated loved one helped this person with their grieving, but the risks are apparent in the story.
The most likely scenario is that these new “intelligent” applications will be a tremendous asset for society, while simultaneously watering the seeds of our destruction. Social media burst on the scene as something mostly fun and definitely addictive. With time, the rot seeped into this online world and brought with it bullying, racism, manipulation, corruption, greed, anger and all those wonderful dark characteristics of humanity. At the same time, it was a place where the forces of love, patience, generosity and joy could build community. Expect no less turmoil in the brave new frontier of artificial intelligence.
Rob Miller is a retired systems engineer, formerly with General Dynamics Canada, who now volunteers with the Calgary Climate Hub and writes on behalf of Eco-Elders for Climate Action. As a climate activist, he works to stop old-growth logging in B.C., reject coal mining on Alberta’s eastern slopes, facilitate community involvement in urban afforestation, and advocate for renewable energy. Miller uses a “systems-thinking” approach to learn, understand, and defend the ecosystems that are under threat by climate change and unrestrained resource development. He lives in Calgary.
Comments
"The answer was remarkably balanced and well-informed. It was not really what I wanted to hear."
The answers were clearly lifted from the copious literature on this subject, providing no new synthesis, just an overview of opinions/journalism. Clever? Yes. Epochal? Not so much.
I asked ChatGPT questions about my area of academic expertise. The answers were factually correct but were clearly a set of strung together sentences and paragraphs. Superficially convincing, but C+ at best; not showing real understanding of underlying phenomena.
In short, although I might use it to slide by in a bird course, I would not trust it to renovate my plumbing!