6/8/25
Will the next generation of humans be widely AI augmented? Could we see new human species split within the next few centuries? You might be surprised…
Where will the future take us? Anatomically modern humans (Homo Sapiens) are thought to have been on Earth for approximately 300,000 years. In the broader sense, the genus Homo - the homenins, or cousin human species with which we could, and did, interbreed - shared many of our characteristics such as symbolic thought. Homenins appear to have existed on earth for about 2.4 million years. But we live now in an age of exponential change, where the changes of the coming centuries are likely to far outstrip those of past centuries. Where does the human story take us from here?
I supplied two top AI models - Gemini 2.5 Pro and Claude 4 Opus - with information about the history of homenins and asked them to extrapolate forward into the future.
Why would a large language model have any credibility at future predictions?
Because they broadly sample human knowledge, AI models can be thought of as a form of compression that extracts the most likely paths from the available information. AIs are trained on the internet (human culture), and that culture itself reflects the swarm intelligence of the 7+ billion members of the human species. When investigating a broad, popular topic such as the future of humanity, a great deal of pre-computation has effectively been done by the crowd and embedded in the data, so we can expect AI compression to be effective at surfacing underlying wisdom.
The table below summarizes the results:
| Time Period | Google Gemini 2.5 Pro Preview | Anthropic Claude 4 Opus |
|---|---|---|
| ~50 Years (2070s) | • Homo sapiens still dominant | |
| • Widespread non-heritable augmentation | ||
| • Neural interfaces common | ||
| • AI as ubiquitous partners | ||
| • Early mind uploading experiments | ||
| • Off-world outposts on Moon/Mars | ||
| • Societal stratification from tech access | • First "Homo augmentus" with permanent neural interfaces | |
| • AI systems matching human physical dexterity | ||
| • Homo sapiens still numerically dominant | ||
| • Blurring boundaries between human and AI | ||
| ~100 Years (2120s) | • Emergence of Homo augmentus | |
| • Germline genetic editing accepted | ||
| • Deep cybernetic integration | ||
| • Advanced AI with potential rights issues | • Three distinct lineages: unmodified sapiens (minority), augmentus, and artificial beings | |
| • First "consciousness transfers" | ||
| • Divergent reproduction methods (biological, technological, hybrid) | ||
| ~200 Years (2220s) | • Continued specialization of human offshoots | |
| • Off-world populations beginning to diverge | ||
| • Digitally native (software only) humans at ~300 years | • Homo cosmicus emerges - adapted for space habitation | |
| • Reproductive barriers between Earth and space humans | ||
| • AI designing own evolutionary paths | ||
| • Deep human-AI consciousness merging | ||
| ~500 Years (2520s) | • Spectrum of interconnected intelligences | |
| • Diverse biological descendants | ||
| • Mature AGIs with possible civilizations | ||
| • Hybrid biological-synthetic beings | ||
| • Potential interstellar diaspora | • Homo genus fragments: aquatus (ocean), cosmicus (space), digitalis (virtual), synthesis (bio-AI hybrids) | |
| • Pure AI lineages diverse as planetary networks and microscopic swarms | ||
| ~1,000 Years (3020s) | • "Hominin" term potentially obsolete | |
| • Extreme diversification and potential speciation | ||
| • Redefined concepts of life and consciousness | • Traditional taxonomy collapses | |
| • Entities can change substrate within lifetime | ||
| • "Species" becomes choice rather than inheritance | ||
| • Multiple-body existence and consciousness merging common | ||
| ~3,000 Years | Not explicitly addressed | • Intelligence inhabiting stars, quantum substrates, new dimensions |
| • Some entities recreate "pure" biological forms as living museums | ||
| ~8,000 Years | Not explicitly addressed | • Archaeological AI studying preserved Homo sapiens |
| • Original humans viewed like we view Homo habilis today |
The full transcripts are available here: Gemini 2.5 Pro Preview, Claude 4 Opus
Both models project a gradual divergence from traditional humanity through augmentation, and genetic modification - and in a faster timeframe than you might expect. For instance, both projected the beginnings of what we might later see as a species split within the next 100-200 years, as humans colonize the solar system and begin operating in different environments. Widespread AI augmentation comes ever earlier - within approximately 100 years. They saw AI expanding greatly, potentially developing its own civilization within ~500 years, while within 1000 years, they expected our traditional taxonomy of species to have collapsed completely, with intelligences being able to changing their substrate - for example, to be born in a body, upload to a digital human form, and from there to merge with AI in hybrid structures we don’t yet have names for. Claude went further, expecting that within approximately 3,000 years intelligence might inhabit substrates far outside currently known capabilities, such as living as patterns of information encoded in stars.
These projections might seem completely fantastic. After all, we have buildings on Earth that are more than 3,000 years old, so that might not seem nearly sufficient time for the next stage of human evolution. There is a saying in forecasting to “never mistake a clear view for a short distance.”
On the other hand, humans think linearly and are notoriously poor at projecting technologically-mediated change - which is exponential, not linear. Technological breakthrough unlocks new capabilities by removing a fundament constraint that limited growth. For example, for most of human history, communication could only happen within the range of a human voice. The invention of long-distance communication opened new possibilities for interaction. With the constraint removed, growth now has no external limitation (ex. vocal range), so is subject only to internal limitations - the rate at which distance communication capabilities can be paid for and deployed. The rate of adoption now resembles an “exponential curve.” More properly, it is really a logistic (”S-curve”) because eventually a new external constraint is reached, such as market saturation, and the adoption rate rolls over.
A new technology that is deployed to take over as the prior one begins to falter can effectively extend the exponential growth. This has happened in computation (aka “Moore’s law”) where four successive generations of technology (mechanical, electromechanical relays, vacuum tubes, and the transistor) have preserved an unbroken exponential scaling.