Song, AI and the Cost of Progress

Technology is bringing us closer to whale intelligence, but at what environmental price?

Dr Cindy Friedman-van der Westhuizen is a lecturer in AI Ethics at Stellenbosch Business School Executive Development

Every year, between June and November, whales return to South Africa’s coastline on their ancient migration from Antarctica to the warmer waters of the Indian Ocean. Their journey predates us. Long before shipping lanes, coastal roads or data centres, whales were already moving through these oceans in rhythms older than memory itself.

For decades, humans could only listen to their gentle singing in awe. In the 1960s, scientists discovered that humpback whales communicate through complex songs, a breakthrough that helped inspire the global “Save the Whales” movement and contributed to protections that pulled several whale populations back from the brink of extinction.

Now, scientists are attempting something even more extraordinary: not just listening to whales but trying to understand them.

The Cetacean Translation Initiative (CETI), a global research project using artificial intelligence and machine learning, is analysing sperm whale clicks to decode their communication patterns. Sperm whales communicate using rapid sequences of clicks known as codas. By analysing thousands of these vocalisations, researchers have identified recurring structures involving rhythm, tempo and variation. The patterns are so sophisticated that scientists have compared them to elements of a phonetic alphabet.

AI has become central to this work. What once took researchers months to analyse manually can now be processed at remarkable speed, allowing AI systems to identify subtle communication patterns humans may never have recognised on their own. The long-term ambition is astonishing, with the eventual interspecies communication possibly in sight.

The idea carries a certain wonder. That technology could bring us closer to nature, rather than further away from it. That AI might help humans reconnect with forms of intelligence we have long ignored.

But beneath this optimism lies a difficult contradiction.

The same AI systems helping us understand whales are also contributing to the environmental crisis and climate change, threatening their survival.

Artificial intelligence is extraordinarily energy-intensive. Training large AI models requires immense computing power, vast amounts of electricity, enormous data centres, water-intensive cooling systems and hardware dependent on environmentally destructive mining practices.

One study suggests that training a single large language model (LLM) generates approximately 300 000 kg of carbon dioxide emissions. This is five times the lifetime emissions of an average car or equivalent to 125 round-trip flights between New York and Beijing.

The energy required to utilise AI is also significant. A single LLM query requires 2.9 watt-hours of electricity, compared with 0.3 watt-hours for a regular internet search.

And the environmental cost does not end once these systems are built. Every AI query, image generation request, automated process and data interaction contributes incrementally to rising energy demand. As AI rapidly expands across nearly every sector of society, so too does its environmental footprint.

Meanwhile, climate change is already reshaping life beneath the ocean’s surface.

Whales are increasingly affected by warming oceans, shifting migration routes, declining food sources and changing ecosystems. Krill populations, the critical food source for many whale species, are declining in some regions as sea ice disappears. Ocean acidification is threatening squid populations that rely on deep-diving whales such as sperm whales. Scientists also warn that warmer waters may increase disease vulnerability and exposure to harmful algal blooms.

In other words, while we are developing technologies to better understand whales, we are simultaneously accelerating the conditions that endanger them.

This does not mean AI should be abandoned. AI holds immense potential for medicine, education, conservation and scientific discovery. It may even deepen our understanding of the natural world in ways previously unimaginable. CETI itself reflects something profoundly human: the desire to connect, to understand, and perhaps to rediscover humility in the presence of another form of intelligence.

But technological progress cannot remain disconnected from environmental responsibility.

There is growing global recognition of this challenge. International initiatives and ethical frameworks are beginning to push for environmentally sustainable AI development, while researchers explore ways to reduce the energy demands of large-scale computing. Yet critics warn that many of these commitments remain largely symbolic, lacking meaningful enforcement or measurable accountability.

And symbolic gestures are something the planet can no longer afford.

Perhaps that is the uncomfortable truth sitting beneath the romance of whale song and technological wonder. We do not merely need better ways to listen to nature. We need the willingness to stop harming it.

If humanity does one day succeed in speaking to whales, perhaps the more important question is not what they are saying to us but what we could possibly say back.