ENCOUNTERS WITH AI VOICES

What are we engaging with when using AI models that generate a podcast episode from sources? What is happening to all the information we are inputting? How is this interpreted, assimilated into patterns, and being unravelled to be constructed in other formats?

Using machine learning has an impact beyond the event of its use. It is a tool. It is a possibility. We use machine learning to get there quicker and with more insight. But what are the potential consequences of using it? What part does human experience play in the process? What do we bring to the table?

More importantly, what is this strange feeling that it creates?

Recently we used the NotebookLM, Google function to generate a podcast episode based on our recent edition of the Field Notes of the Biota Beings. Two AI voices discuss the work, they elaborate on the writing. Their tone is enthusiastic and engaged. It is striking how naturalistic the voices sound. After listening for a while, the voices can persuade the listener they are human. Before these tools become second nature, we want to reflect on how we change as we use them.

There are lots of positives of listening to the AI generated discussion. For instance, it pinpoints development opportunities and helps see the work from a distance. The AI generates interesting points of clarification and ‘sound bytes’ from a third ‘person’ perspective. Here are some quotes from the AI discussion:

Voice 1: “The Biota Beings are really hinting that we have lost that sense of reverence.”   

Voice 2: “Yeah.”

Voice 1: “That understanding that we are not separate from nature.”

Voice 2: “So, it’s not about building cool eco-friendly apartments for squirrels. It’s about shifting our whole mindset.”

Voice 1: “Exactly… and they get super-specific about what this shift looks like.”

Later in the discussion, when reflecting on the creative prompt aspect of the Field Notes:

Voice 2: “Less rulebook, and more choose your own bio-adventure.”

The AI generated podcast presents a discussion between two voices, one intentionally sounding male and the other female. Here we discover one of the potential points of contention. The AI voices present their conversation as a discussion, but it isn’t a true discussion. This is a fabricated output created by a singular form and presented as a discussion between multiple people.

“Quite quickly you stop listening to AI voices. And you start to think about them as people.” [quote taken from our analysis of the experience]

There also appears to be limited criticality in the reflection on the source material. It delivers information as a point of truth. In the small print below the Chat screen, it reads “NotebookLM can make mistakes, so double-check it” - but the conversation is compelling.

We question whether hearing other voices talk about the work in a particular way taps into a social instinctive cue. Our social instincts lead us to “follow the herd.” Could the discussion for the unaware appear to create social proof?

As one of the co-creators of the work, there was another reaction taking place whilst listening. One that could have consequences if not realised.

In using the AI discussion function, hearing the conversation, its positive tone of voice, inspires the feeling of confidence in the work. It is persuasive. The social proof that the AI discussion implies lacks some criticality. It begins a feedback experience to build reassurance in the work. Even though we are aware of the AI source, included in the minor glitches in the voice simulation, the voices feel human. Despite the reminder of the artificiality, however, the experience inspires a below-conscious reaction, as if it was from human sources.

“It’s amazing, I’m impressed.”

“Yeah, I’d have to be careful not to use it [AI generated podcast content] too soon, and then in my head it feels it’s finished. Because it presents it in such a finished product. .. I would miss the ciritical [development].”

[quotes taken from our analysis of the experience]

When analysing the experience and our use of the AI function, there lurks a feeling of unease. There is an opportunity to encounter some unknown consequences of these tools. Some of these may be to further generate echo chambers and information bubbles. Misplaced confidence in a work may not encourage development and questioning. Perhaps confidence is an important strength to have, but there is also a time for critique and analysis.

“If we are not thinking critically, we will just absorb this information.”

[quote taken from our analysis of the experience]

Perhaps there are lessons to be had in response to these growing sets of tools. It calls for a personal awareness of strengths and weaknesses. Overconfident individuals may interpret this as an assurance of their ‘rightness’ and be less open to challenging existing knowledge. For the uncertain, it could provide reassurance. For the biased, it might reaffirm points of view and create echo chambers.

From the initial use of the tools, there appears to be a simplification of ideas to a relatable and actionable interpretation of the source material. In its current state, the AI resists abstract extrapolation.

There is no doubt that the AI tools are incredibly useful, powerful and world changing. However, from these early engagements, there is a requirement to be aware of our human pitfalls and a potential growing reliance on these tools. 

Just maybe, there is a time to engage in other forms of creative prompts and human experiences alongside the growing AI tools.  For these occasions, tools such as the Field Notes of the Biota Beings, can be a companion on journeys of discovery.

To access the AI generated episode mentioned in this article please visit the recording our our YouTube channel at: https://youtu.be/X3UV3HPU0z4

Purchase your copy of the Field Notes of the Biota Beings: Forest Fields here.