The limiting factor in ocean data intelligence is no longer the technology. It is accessibility.

Edge AI, autonomous navigation, miniaturized multi-modal sensing, satellite connectivity, cloud infrastructure, and advanced manufacturing have all crossed the thresholds that once kept ocean intelligence the exclusive province of major institutions. We can go deeper, stay longer, see more, and process faster than at any point in human history. And yet the people, organizations, and communities who most need ocean data (conservation NGOs, developing-nation fisheries managers, coastal communities, citizen scientists, small offshore operators) largely cannot access it. The tools exist. They remain designed for, priced for, and gated by institutions, not the broader communities who need them most.

This is the central problem of ocean intelligence today. It is not a capability problem. It is an access problem: in who the tools are designed for, what they cost, and what they tell you. It runs through every other challenge in ocean data collection: the hostile operating environment that drives costs up, the coverage gaps that persist because the tools are too expensive to deploy at scale, the processing bottlenecks that grow as collection outpaces analysis, the fragmentation that keeps what does get collected siloed and unusable. Accessibility is the constraint that connects them. It is also the one we are now, for the first time, equipped to solve.

This framing is not new. A 2025 Science review of the UN Ocean Decade argued that “accessible, cost-effective technologies must be developed and shared widely, lowering technological barriers to near zero to ensure equitable global access.”¹ The international consortium coordinating ocean observation (the Marine Technology Society, the Global Ocean Observing System, NOAA, and industry partners) has named the same priority.² The third Dialogues with Industry, held in late 2025, was convened explicitly to “develop actionable pathways that the Ocean Enterprise can implement to advance the democratization of ocean observation.”³ The diagnosis is widely shared.

Why This Moment Is Different

The ocean’s capacity to buffer human activity is not infinite, and the evidence that it is reaching its limits is no longer speculative. The rate of ocean warming has doubled over the past two decades, as documented in the IOC’s 2024 State of the Ocean Report. The ocean’s role as a carbon sink, one of the most important mechanisms slowing the pace of climate change, shows measurable signs of weakening and, beyond certain thresholds, could become self-reinforcing. The evidence that the ocean is changing is no longer in question. The questions now are how fast, in what directions, and what can still be done about it.

Meanwhile, the demands we place on the ocean are intensifying. Offshore wind is scaling rapidly; aquaculture is expanding to meet growing protein demand; wild capture fisheries face mounting pressure from overfishing, habitat degradation, and climate-driven species shifts; and deep-sea mining is moving from concept to permitting. Each of these activities generates a need for data: environmental baselines, real-time monitoring, impact assessment, regulatory compliance, risk management. And in nearly every case, the data infrastructure is inadequate, not because the sensing technology does not exist, but because it has not been made available at the scale, cost, or usability the situation requires.

The Challenges

The problems that limit ocean data collection are real, persistent, and not reducible to a single cause. They span the physical, the economic, the technical, and the institutional. Each is worth understanding on its own terms, because each shapes what is possible and what is not.

The ocean is hostile to technology

The underwater environment presents engineering challenges that have no parallel on land. Pressure increases by one atmosphere for every ten meters of depth. Saltwater corrodes metals, degrades seals, and attacks electronics. Biofouling, the colonization of surfaces by marine organisms, can obscure a camera lens or clog an intake within weeks. Light attenuates rapidly; below a few dozen meters, vision-based systems must contend with low light, particulate scatter, and limited range. These conditions demand fundamentally different approaches than their terrestrial equivalents. GPS signals do not penetrate water. Acoustic communication, the primary alternative, is low-bandwidth, noisy, and range-limited. Marine life can damage equipment through predation, collision, or entanglement.

These are ordinary operating conditions, and every piece of hardware that works underwater carries the full cost of engineering against them: in materials, in sealing, in redundancy, in testing. That cost is one reason the tools that succeed tend to be expensive, specialized, and designed for users who can maintain them. A 2024 review in ACS Sensors describes how biofouling and corrosion remain dominant constraints on marine sensor design, limiting deployment durations, requiring active wiper or coating systems, and adding to the engineering burden of any instrument intended to function long-term in seawater.⁴

This shows up everywhere. Offshore wind monitoring requires instruments that survive continuous exposure to North Sea and Atlantic conditions. Subsea cable inspection covers thousands of kilometers of deep, remote seabed. Coral reef surveys demand equipment that functions in biofouling-prone tropical waters with complex three-dimensional topology. And any imaging system, across any of these applications, needs housings that resist algae growth on the optical window, or it eventually goes blind. The environment does not make exceptions.

We are not watching enough

The ocean is dynamic at every timescale (tidal, diurnal, seasonal, interannual, decadal) and across every spatial scale, from reef patch to ocean basin. Our monitoring of it is sparse, episodic, and overwhelmingly stationary. The result is a patchwork of data with gaps in baselines, in event coverage, in seasonal patterns, and in long-term trends.

The scale of the gap is well documented. A PLOS ONE global analysis of cetacean line-transect surveys found that less than 25 percent of the world’s ocean surface had been surveyed at all, and only 6 percent had been covered frequently enough to allow trend estimation⁵. Cetaceans are among the more closely monitored marine taxa. The picture is comparable across other groups. Coral reefs are often surveyed only once or twice a year, if at all. Bleaching events can begin and end between visits, leaving their onset, progression, and recovery undocumented. Fisheries stock assessments cover fractions of a species’ range and are often years behind the reality they are meant to describe. Marine protected areas are designated based on ecological promise, then left without the monitoring infrastructure to evaluate whether the designation is working. Offshore wind farms spanning tens of square kilometers may be monitored by a handful of stationary buoys, each sampling a fixed point in a system defined by movement.

The trend is not improving. The IOC’s 2024 State of the Ocean Report found that, despite rapidly expanding demand for ocean information, there has been no significant increase in sustained global ocean observations over the past five years.⁶ Demand is rising faster than sustained observing capacity. That gap is the substance of the problem. That gap is the substance of the problem.

The ocean does not hold still between survey visits. Temporal resolution matters as much as spatial coverage, and we are falling short on both.

The cost barrier

The tools that meet research-grade requirements remain expensive to deploy at the scale the problems demand. Research vessels cost between $4 million and $10 million per year to operate. Commercial autonomous underwater vehicles range from $50,000 to $500,000 per unit. Remotely operated vehicle inspection runs upward of $10,000 per day. UNESCO’s Global Ocean Science Report 2020 inventories these costs and the broader infrastructure of ocean science worldwide, and finds that, on average, only 1.7 percent of national research budgets is allocated to ocean science, a figure that constrains both the building and the operation of the platforms the field depends on.⁷ These figures are not incidental to the coverage gaps described above. They are the mechanism by which those gaps persist.

The cost barrier is not distributed evenly. The coral reefs most at risk from climate stress are disproportionately located in developing nations with the least monitoring infrastructure. Conservation NGOs operate on budgets that cannot absorb the cost of persistent monitoring. Small offshore operators (marinas, dive businesses, coastal aquaculture farms) have practical need for environmental data but no realistic path to the instruments that produce it. Citizen science communities possess the motivation, the geographic reach, and often the domain knowledge, but not the budgets.

The result is a system in which the ability to collect ocean data is concentrated among the institutions that can afford the tools, while the communities closest to the ocean, and most affected by its changes, are the least equipped to observe it.

Collection outpaces analysis

Even where data is collected, the volume frequently overwhelms the capacity to process it. Passive acoustic monitoring arrays can generate gigabytes of audio per day, and large observatory networks produce hundreds of gigabytes daily, much of it sitting in backlogs awaiting analysis. Seafloor imaging campaigns produce terabytes of visual data that take months or years to review. Satellite imagery for maritime enforcement requires analysts to scan thousands of images for signs of illegal fishing activity, and optical satellites are frequently obscured by cloud cover. As a 2022 study in Ecology and Evolution put it, even in terrestrial bioacoustics, passive acoustic monitoring is an area where “technological advances have facilitated an influx of data that routinely exceeds the capacity for analysis.”⁸ The same pattern holds across ocean sensing modalities.

The problem is compounded by inconsistency. Different institutions use different instruments, different calibration standards, different metadata schemas, and different file formats. Data collected by one program often cannot be compared or combined with data collected by another, even when they are measuring the same parameters in the same region. The result is not just a processing bottleneck but a fragmentation problem: data that exists in quantity but cannot be aggregated into the kind of coherent, multi-source intelligence the decisions demand.

What we collect stays siloed

The data fragmentation problem runs deeper than format incompatibility. It is institutional, structural, and in many cases, intentional. Commercial marine industries (offshore energy, shipping, aquaculture, cable maintenance) collect substantial quantities of environmental data in the course of their operations. The vast majority of it remains proprietary, never entering the shared repositories where it could inform conservation, research, or regulatory decisions. A 2019 Frontiers in Marine Science review of ocean data services found that, despite the existence of FAIR data principles and substantial investment in interoperability infrastructure, “true international interoperability seems only to have been achieved for a small fraction of the kinds of data being collected in the world’s oceans.”⁹

Academic research faces a different version of the same problem. Grants typically fund one sensor for one mission. A marine biologist deploying a hydrophone array to study fish behavior, for example, may have no funding or mandate to add even a basic CTD sensor to the same platform, even though the opportunity to collect co-located environmental context will not recur once the instrument is retrieved. Every single-modality deployment is a missed opportunity for multi-modal intelligence: a sound without a visual, a temperature without a current, a snapshot without a baseline.

Cross-border environments add another layer. Fisheries that migrate across national boundaries are managed by nations with incompatible data standards, collection methods, and sharing agreements. The ocean does not observe jurisdictional lines. Our data systems do.¹⁰

Five problems: overlapping in cause, bound by one constraint. Until recently, no available technology could resolve them at the price points and form factors that accessibility requires. What is new is twofold: the technologies that make accessible ocean intelligence feasible have matured, and the design choice to build systems around accessibility, rather than scaling down from institutional ones, can now be made.

The Convergence

What has changed is not any single enabling technology. It is that all of them have matured and become affordable at roughly the same time, a convergence that took shape over the early 2020s.

The pattern is recognizable from another domain. In 2013, the DJI Phantom shipped at $629: the first consumer quadcopter that worked out of the box, the moment aerial imaging stopped being the exclusive province of film crews and aerospace contractors. Within a decade, drones were being used for precision agriculture, infrastructure inspection, disaster damage assessment, conservation monitoring, and search and rescue: applications that the platforms’ designers did not anticipate. None of the enabling technologies (GPS, brushless motors, lithium batteries, image stabilization) were new in 2013. What was new was that they had all crossed the same thresholds at once (small enough, cheap enough, and available off-the-shelf) to be combined in a consumer product. The same dynamic has since extended into domains the original designers explicitly disclaimed: a 2024 analysis of the Russo-Ukrainian war argues that off-the-shelf consumer drones from manufacturers like DJI and Autel have reshaped the practices and aesthetics of modern warfare, with “wedding drones” repurposed by volunteers becoming central to frontline operations.¹¹ Whatever one thinks of that turn, it makes the point: when capable hardware becomes broadly accessible, the applications that follow exceed what any designer can plan for.

Ocean intelligence is not just at the same threshold; it is building on it. Many of the components that made consumer drones possible (flight controllers, IMUs, brushless motors, lithium chemistries, low-power telemetry) are now available because the drone market drove them there.

Edge AI hardware (processors capable of running real-time inference models onboard an underwater platform) now costs hundreds of dollars, not tens of thousands. A decade ago, onboard intelligence meant a satellite uplink and a shore-based processing center. Today, it means a module the size of a credit card that can classify species, detect anomalies, and make navigation decisions without surfacing.

Multi-modal sensing has undergone a parallel cost collapse. Cameras, hydrophones, CTDs, and inertial measurement units that were once individually expensive enough to anchor an entire instrument budget can now be combined on a single platform at a fraction of the historical cost. Collecting visual, acoustic, and environmental data simultaneously (and fusing them into something more informative than any single modality) is no longer a laboratory capability. It is an engineering choice.¹²

Advanced manufacturing and materials science have lowered the barrier to producing marine-grade hardware. Composites, 3D-printed tooling, and modular design architectures allow for robust, repairable, adaptable platforms without the capital expenditures that once made marine hardware development the exclusive province of defense contractors and research institutions.

And data infrastructure has caught up with the volume the ocean produces. Cloud storage is cheap; processing is elastic; visualization, mapping, and data-serving tools are mature commodity services. Satellite networks like Iridium, Swarm, and Starlink make real-time telemetry from remote ocean locations feasible at price points that would have been unimaginable a generation ago.

None of these technologies is individually new. What is new is their simultaneous maturity, and the design space they collectively open: building ocean intelligence systems around accessibility rather than institutional capability.

The drone pattern is not unique. When GPS moved from military positioning to consumer navigation, the applications that followed (turn-by-turn driving directions, precision agriculture, ridesharing) were not predictable from the original technology. When inexpensive weather stations reached hobbyists, Weather Underground built a network of over 250,000 stations that now fills coverage gaps in the official meteorological network. Volunteer divers trained through Reef Check and Reef Life Survey produce biodiversity data of sufficient quality to appear in peer-reviewed scientific literature.

Ocean intelligence is the next domain in which this pattern can play out. The applications that emerge when the tools reach fisheries managers, reef conservationists, offshore operators, coastal communities, and citizen scientists will exceed what any platform designer can anticipate. The value is not just in better data; it is in the entirely new categories of data, application, and insight that become possible when the barrier to participation drops.

What Comes Next

If accessibility is the constraint that appears in every problem, the response is the convergence of two things: mature technologies that are now affordable enough to deploy at scale, and design choices that put those technologies in the hands of the people who need them.

Hardware built for the ocean from the outset: answering the hostile environment and the cost it imposes. Not adapted from terrestrial platforms, not scaled down from institutional systems, but conceived for the environment it operates in. Deployable by hand, from a small boat or a dock. If it requires a research vessel, a crane, and a technical crew to put it in the water, it is not accessible. Biomimetic approaches, borrowing design principles from organisms evolved for the underwater environment, offer a path to hardware that is efficient, robust, and natively adapted to the medium.

Embedded intelligence, not offloaded processing: answering the collection-versus-analysis bottleneck. AI should do the hard work (navigation, pattern recognition, anomaly detection, data triage) so that the operator sets the mission and the platform executes it. Mission planning should be legible to someone without an engineering background. The intelligence exists to close the gap between the ocean’s complexity and the user’s need for clarity.

Legible data, not raw output: answering the fragmentation and processing problems together. Answers, not data dumps. Coral health status at a specific site. Species presence or absence. An anomaly at specific coordinates requiring attention. If understanding the output requires a marine acoustics degree, the system has not finished its job.

Built for anyone with a reason to care: answering the coverage gaps that institutional tools cannot close. The marine biologist, certainly. But also the dive operator tracking reef health for a local conservation group. The fisheries manager in a Pacific island nation enforcing a marine protected area with a two-person team. The coastal community monitoring water quality after a storm. The citizen scientist contributing observations to a biodiversity database. Urgency does not check credentials, and the tools should not either. A 2024 study in Conservation Letters comparing volunteer scuba diver observations with a federally-funded NOAA fish population survey over 25 years concluded that citizen scientists “can be effective sentinels of ecological change,” with substantial value in monitoring otherwise data-limited marine species.¹³ Trained dive practitioners, volunteer monitors, and citizen scientists already produce data that appears in peer-reviewed literature when the tools support them.

The challenges in collecting ocean data (hostile conditions, coverage gaps, prohibitive cost, processing bottlenecks, institutional fragmentation) are real and persistent. They will not be solved by any single technology or any single organization. But accessibility, the most pervasive of those challenges, is now the most solvable, because the technologies that enable it have converged.

The question is no longer whether ocean intelligence at scale is technically possible. It is. The question is whether we build the tools that put it in the hands of those who need it, and build them now, when the decisions that will shape the ocean’s future are actively being made.

We need ocean intelligence, now. The technology exists. The urgency exists. What remains is to build the tools that connect them.

References

1. Miloslavich, P., et al. (2025). The United Nations Ocean Decade: A catalyst for international collaboration enhancing ocean monitoring and data. Science, 390, eaed3111. DOI: 10.1126/science.aed3111

2. Lindstrom, E., Gunn, J., Fischer, A., McCurdy, A., & Glover, L.K. (2012). A Framework for Ocean Observing. Task Team for an Integrated Framework for Sustained Ocean Observing, Paris, UNESCO/IOC. DOI: 10.5270/OceanObs09-FOO

3. Fietzek, P., Kocak, D., Zinkann, A.C., Willis, Z., & Closek, C.J.E. (2026). Dialogues with Industry - Ocean Observing of the Future: Sensors, Instruments, Platforms and the Market Ahead. Background Paper. Marine Technology Society / Global Ocean Observing System / NOAA / IOOS (Report no. MTS-202506).

4. Sahoo, B.N. (2025). Antibiofouling Coatings For Marine Sensors: Progress and Perspectives on Materials, Methods, Impacts, and Field Trial Studies. ACS Sensors, 10(3), 1600. DOI: 10.1021/acssensors.4c02670

5. Kaschner, K., et al. (2012). Global Coverage of Cetacean Line-Transect Surveys: Status Quo, Data Gaps and Future Challenges. PLOS ONE, 7(9), e44075. DOI: 10.1371/journal.pone.0044075

6. IOC-UNESCO (2024). State of the Ocean Report 2024. Paris, IOC-UNESCO. (IOC Technical Series, 190). DOI: 10.25607/4wbg-d349

7. IOC-UNESCO (2020). Global Ocean Science Report 2020: Charting Capacity for Ocean Sustainability. K. Isensee (ed.), Paris, UNESCO Publishing.

8. Symes, L.B., et al. (2022). Analytical approaches for evaluating passive acoustic monitoring data: A case study of avian vocalizations. Ecology and Evolution, 12(4), e8797. DOI: 10.1002/ece3.8797

9. Tanhua, T., et al. (2019). Ocean FAIR Data Services. Frontiers in Marine Science, 6, 440. DOI: 10.3389/fmars.2019.00440

10. Muller-Karger, F.E., et al. (2023). Marine Life 2030: Building global knowledge of marine life for local action in the Ocean Decade. ICES Journal of Marine Science, 80(2), 355-357. DOI: 10.1093/icesjms/fsac084

11. Bender, H., & Kanderske, M. (2024). Consumer Drone Warfare: Practices, Aesthetics and Discourses of Consumer Drones in the Russo-Ukrainian War. In E. Serafinelli (Ed.), Drones in Society: New Visual Aesthetics. Springer Nature. DOI: 10.1007/978-3-031-56984-5_11

12. Claustre, H., Johnson, K.S., & Takeshita, Y. (2020). Observing the Global Ocean with Biogeochemical-Argo. Annual Review of Marine Science, 12, 23-48. DOI: 10.1146/annurev-marine-010419-010956

13. Greenberg, D.A., Pattengill-Semmens, C.V., & Semmens, B.X. (2024). Assessing the value of citizen scientist observations in tracking the abundance of marine fishes. Conservation Letters, 17, e13009. DOI: 10.1111/conl.13009

Long awaited Artificial Intelligence is here with Artificial General Intelligence (AGI) matching human intelligence not far behind. Like a Moore’s Law script, fulfilling a Singularity prophecy, surpassing human creativity, by 2045. (Kurzweil, 2024). And with this technological wave, there will be incredible consequences along with its incredible benefits. The forecast is that AI will have a profound affect on contemporary life; from personal and business management to research and education to political and social power to health and human services. Back in 2020, Peter Diamandis, co-founder of the X Prize Foundation and Singularity University, predicted the next decade we’ll experience more technological progress than in the past 100 years (The Guardian, 2020) and his friend and futurist Ray Kurzweil has long predicted by 2045 the intelligence Singularity (Popular Mechanics ) which would bring an explosion of technological advancement bringing “abundance for all”. While futurists praise the coming technologies that will bring us abundance and livelihood, we are now weathering the AI tempest, feeling it overtaking us in creative power, outpacing society and government’s ability to adapt and driving us into exciting yet potentially dangerous waters before finding this supposed oasis.

No matter how AI evolves, it is certain that children now will grow up in an environment of decisions and creativity not just guided but dictated by artificial systems. It is imperative, therefore, that we equip this Generation AI with the skills and ethical understanding necessary to maintain their humanity against the waves of innovation and technological breakthroughs driven by machines built in the unforgiving world of capitalism.

This concern was reinforced after an attendance to Columbia University’s FabLearn conference focused on the “Constructionism” pedagogy in education. Constructionism, like Project Based Learning (PBL), Papert believes learning is facilitated by constructing actual artifacts or objects - whether a theory, a sandcastle, or a computer program - which can then be shared and discussed with others (EBSCO). In other words, do instead of being told. Work through problems in real time, learning by doing. By encouraging students to engage in real-world projects and solve authentic problems, they build knowledge through practical, hands-on experiences and develop intellectual “muscles” that require practice. Parallel to this pedagogical approach, the “Design Process” (Interaction Design Foundation) has become a significant tool prioritizing the needs and experiences of the end users during the creation process. Embedded within this creation process is an essential fact-finding exercise, labeled “Emphasize” aimed at understanding the user and their specific challenges: Empathizing with the user is crucial as it leads to a deeper comprehension of the problem at hand, enabling the design of more effective and tailored solutions. At this conference about the physical act of creation, within nearly every talk and workshop and indeed now throughout the education world, the topic of Artificial Intelligence (AI) and it’s affects on education were being discussed: how it can guide or even dictate not only the learning process but also the act of building and creation itself? For better or for worse, how could it short circuit the learning process in nearly every subject. How will this transformative tool affect future creators who may no longer need to directly engage in a discovery process with the user, reader, or viewer for which their creations are to be made? What happens when we don’t exercise the power of Empathy?

The introduction of ChatGPT and LLM’s (Large Language Models) has set the stage for the next generation to grow up with (theoretically) all of human knowledge at their fingertips, intimately interacting with and influenced by AI from womb to coffin. This shift in access to knowledge profoundly alters the dynamic between creator and the user in the Design Process, fundamentally reshaping the very nature of how we form ideas about the problem to be answered. Additionally, there are limitations of this knowledge creating an inherent bias within the informer. LLM’s are based on our language after all. Our digital texts from books, magazines and blogs not only contain memories, our knowledge and our history but also dreams, fears and beliefs - our biases. However, it certainly doesn’t include every text nor every perspective or every story. Like humans, there is a bias built from a lifetime of experiences as well as a lack of knowledge and experience. “Science” in fact was created to overcome these biases to we can make informed decisions and interpretations of data. AI trained on text selected by a capitalistic, privileged minority will undoubtedly reinforce the biases of its teachers. And right now, millions of people including young students are trusting the AI output, as the human counterpart, users released from the responsibility of being critical and having an agreeable echo chamber of ideas about and for the world, leveraging the immense shortcuts in the creative process. But at what cost to these young minds?

In Constructionism, the act of exploration of potential solutions both empowers the students and challenges them to dig deep into fundamental exercises of critical thinking while opening up the channels of creativity. This learning process, particularly when building and designing something, requires practice in process. It’s a muscle that needs to be exercised, in danger of atrophying. No longer required to ask or investigate, no longer required to work through personal biases to understand your neighbor whose life experiences are different than yours, now dependent on a machine to tell you how others feel and what they think, would AI then become a crutch in the creation process, an emotionless translator for human emotion? (Inzlicht, 2024) ) Surely tools like ChatGPT increases innovation and iteration speed, right? But at what cost does it have to the nurtured skill of empathy?

With easy access to the knowledge of all humanity’s experience, it’s theoretically democratizing to have such information at your fingertips. But in the process of making ideas more accessible, what will such power do to the skills that have been required of us humans to investigate and question in order to achieve such creativity? Will quick and easy answers deprive our children needed lessons that foster the skill of Empathy? And will the adults they become lack the Empathy we desperately need to live in harmony with one another in this world that is getting ever smaller and ever more in danger? We must prepare, by training or teaching, our offspring (both silicon and carbon based),for this Brave New World. Generation AI will need more than ever to be taught and reminded of what it means to be human, what it means to always to attempt to understand and appreciate your fellow humans and creatures of this Earth. This includes fostering critical thinking, emotional intelligence, and a robust understanding of AI ethics and governance, ensuring they can make informed decisions and maintain human agency in a world increasingly shaped by non-human intelligence. Our stories and beliefs around AI certainly makes it seem we are at this fateful spark, potentially igniting a revolution that could lead to our salvation … or heralding our prophetic doom. To avoid succumbing to the mortal temptations to whom (or what) we surrender power and loose ourselves along the way, we will need to build a guiding light for Generation AI caught in potentially unforgiving artificial intelligence storm.

References

Corbyn, Zoë. (2020, January 25). Peter Diamandis: “In the next 10 years, we’ll reinvent every industry”. The Guardian. https://www.theguardian.com/technology/2020/jan/25/peter-diamandis-future-faster-think-interview-ai-industry

Orf, Darren. (2025, June 30). A Scientist Says Humans Will Reach the Singularity Within 20 Years. Popular Mechanics. https://www.popularmechanics.com/science/a65253231/2045-singularity-ray-kurzweil-prediction/

Kurzweil, R. (2024). *The Singularity Is Nearer: When We Merge with AI.* New York: Viking.

Inzlicht, M. (Oct 2024). In praise of empathic AI. Rotman Insights Hub, Rotman School of Management, University of Toronto. https://www-2.rotman.utoronto.ca/insightshub/ai-analytics-big-data/empathetic-AI