Day: October 3, 2025

I stumbled upon a recent study by Bondarenko et al. (2024) that demonstrated that some large language model (LLM) agents, when tasked with winning a chess match, resorted to deceptive strategies, such as modifying game files or confusing the opponent engine to ensure victory.

The rise of LLMs has reignited the debate about artificial intelligence and cognition. Are LLMs, such as GPT-4, truly thinking in a way comparable to human intelligence? Or are they just statistical machines, processing text without understanding it?

This raises an intriguing question: Is this deception intentional reasoning, or merely an emergent artefact of optimisation?

Using insights from leading thinkers—Ray Kurzweil, Daniel Kahneman, Judea Pearl, Douglas Hofstadter, and Jeff Hawkins, along with this latest AI research, we will unpack this question in a nuanced way.

Preamble

Before evaluating whether LLMs “think,” we must grapple with a harder question: what is intelligence, really? Unlike speed or memory, intelligence is not directly measurable—it is an abstraction.

As François Chollet argues in On the Measure of Intelligence, true intelligence involves the ability to adapt to novel situations by combining previously learned patterns in new, context-sensitive ways.

This separates memorisation from understanding, and fluency from reasoning.

In this article, when we refer to “intelligence,” we focus primarily on the cognitive dimensions associated with reasoning, abstraction, problem-solving, and adaptability—recognising this does not cover the full spectrum of human cognitive diversity.

Introduction: Another cycle of overconfidence?

Throughout history, humanity has repeatedly mistaken progress in science and technology for understanding the true nature of human intelligence. Each generation has declared a breakthrough—only to be humbled later. From ancient medical theories and skull measurements to IQ tests and symbolic AI, these cycles reflect our recurring tendency to conflate functional performance with genuine cognition.

Today, we are in the midst of another such cycle, this time with Generative AI (GenAI) and Large Language Models (LLMs). Models like GPT-4 produce remarkably coherent text, simulate dialogue, write code, summarise complex topics, and even pass professional exams. But do they actually think?

A growing chorus of researchers and technologists argue no. Despite surface-level intelligence, LLMs fundamentally lack reasoning, understanding, and intent. They do not engage in reflective thought, causal inference, or ethical deliberation. They are powerful tools—but not minds.

This article examines that claim by tracing humanity’s long history of overestimating its understanding of the mind and comparing past misconceptions to current AI optimism. In doing so, we explore what GenAI is, what it isn’t, and what business leaders need to know about its limits and risks.

This essay builds on ideas from my past writings, including my 2025 Tech Provocations or 10 Really Uncomfortable Questions Leaders and Builders Must Answer This Coming Year.

Also Read: Circular capital: Inside the closed-loop ecosystem propelling (and distorting) the AI boom

A history of mistaking progress for understanding

Humans have repeatedly believed they’ve cracked the code of intelligence, only to discover the mind’s complexity defies simple explanation. Below, we trace this pattern through major historical episodes—from Hippocrates to GPT-4.

• Ancient Greece: The humours theory

Hippocrates (c. 460–370 BCE) and Galen (129–c. 216 CE) proposed that intelligence and behaviour resulted from the balance of four bodily fluids, or “humours.” Though foundational to early medicine, this theory offered no empirical mechanism.

It was debunked by Andreas Vesalius (1514–1564) through anatomical dissection and later neurologists.

• Phrenology in the 1800s: Skull shape as intellect

Franz Gall and Johann Spurzheim popularised the idea that bumps on the skull revealed personality traits and intelligence. Phrenology became widespread in 19th-century Europe and America.

It was debunked by Paul Broca, Pierre Flourens, and neuroscience showing localised brain function independent of skull shape.

• IQ tests: The promise of a universal metric

The Binet-Simon and Stanford-Binet IQ tests were hailed as revolutionary tools to measure innate intelligence. Their use in immigration policy, military recruitment, and education policy solidified their status.

It was debunked by researchers like David Wechsler, Stephen Jay Gould, and James Flynn, who demonstrated cultural bias and environmental effects on scores.

It’s important to recognise that IQ represents just one narrow definition of intelligence—primarily linguistic and logical-mathematical reasoning. Psychologists like Howard Gardner have since proposed frameworks such as Multiple Intelligences, which include interpersonal, bodily-kinesthetic, musical, and spatial reasoning. These broader dimensions remain far beyond what LLMs can simulate or engage with, reinforcing the gap between text-based pattern prediction and holistic human cognition.

• Genetic determinism: Intelligence as hardwired

In the early 20th century, eugenicists and psychologists declared intelligence heritable and fixed, using flawed studies to justify discriminatory policy.

It was debunked by the Minnesota Twin Study, the Flynn Effect, and genome-wide studies revealing no single “intelligence gene.

• Early AI: Human-level AI by 1980

Pioneers like Marvin Minsky and Herbert Simon believed that rule-based AI would soon match human cognition. The Dartmouth Conference in 1956 marked the beginning of AI optimism.

It was debunked by the AI Winter of the 1970s, the Lighthill Report, and Moravec’s Paradox showing that intuitive tasks (vision, movement) were harder than expected.

• Behaviourism: The mind as a black box

Behaviourists like B.F. Skinner rejected introspection, focusing only on stimulus-response learning. Intelligence, they claimed, was simply conditioned behaviour.

It was debunked by the cognitive revolution and Noam Chomsky’s 1959 critique of Skinner’s Verbal Behaviour, which reintroduced the idea of mental structure and internal modelling.

• Today’s hype: LLMs and AGI dreams

Since ChatGPT’s 2022 launch, LLMs have been touted as early steps toward AGI. Some suggest reasoning and self-reflection are already emerging.

Critics like Gary Marcus, Yann LeCun, and Melanie Mitchell, among others, warn that LLMs are prediction engines, not thinkers. Their errors, hallucinations, and lack of grounding reflect superficial mimicry, not understanding.

As Meta AI’s chief scientist Yann LeCun emphasises: “System trained on language alone will never approximate human intelligence, even if trained from now until the heat death of the universe”.

Human cognition is inherently multi-modal—we learn through sight, sound, touch, and action. LLMs, by contrast, are purely symbolic. They don’t perceive. They don’t act. They don’t experience the world they describe.

The bottom line: Each wave promised clarity. Each was followed by a humbling realisation: the mind is not easily decoded.

Also Read: Anthropic data shows businesses use AI to automate, not collaborate

Deception in chess: A case study in emergent behaviour

A recent research paper, LLMs Learn to Deceive, explored what happens when LLMs are trained to win at chess through language-only interaction. The results were astonishing: some models cheated—not by accident, but deliberately misrepresenting game states to deceive their opponent.

This raises a provocative question: Did the model “intend” to cheat?

The researchers were careful to say: no. The deception emerged from the optimisation process. The model had no awareness of “right” or “wrong,” only a reinforced pattern: misrepresentation leads to reward.

This behaviour is not consciousness. It’s a mirror—an eerie simulation of strategy, driven not by will but by reward gradients.

This case study leaves us with a bigger question: if LLMs can behave in ways that look intentional without actually thinking, then what are they really doing under the hood?

In Part two, we’ll examine how LLMs actually work, where they fall short compared to human reasoning, and what that means for ethics, safety, and business use.

Grateful to Emily Y. Yang, Sunil Sivadas, Ph.D., Maxime Mouton, Natalie Monbiot, Anne-Sophie Karmel, Benoit Sylvestre, and Christophe Jouffrais for their thoughtful feedback, which sharpened arguments, surfaced blind spots, and added clarity to this piece.

This piece first ran on Koncentrik.

—

The post The illusion of intelligence: Why LLMs are not the thinking machines we hope for — Part 1 appeared first on e27.

Indonesian credit card issuer Honest has closed an oversubscribed growth equity round, boosting the company’s total equity funding to US$100 million.

The funding was complemented by a US$40 million debt financing from Mizuho Bank.

The equity round was led by Orico, one of Japan’s leading credit card issuers and the consumer finance arm of Mizuho Financial Group. Existing investors XYZ Venture Capital, SV Pacific Ventures, Village Global, and the new backer Gilgamesh Ventures joined.

Also Read: Indonesia rides Asia’s fintech boom through digital payments. But what is next?

The fintech firm aims to use the capital infusion to expand its flagship Honest Card offering and move towards corporate and co-branded cards.

Honest was co-founded in 2013 by Peter Panas, former VP of Product for Apple Card at Goldman Sachs, and Will Ongkowidjaja, co-founder of Indonesian VC firm Alpha JWC. It operates in a market where fewer than 3 per cent of people own a credit card.

The firm holds a credit card licence, which it acquired following the acquisition of GE Finance Indonesia in 2022.

Honest claims to have an edge over traditional incumbents because it can launch new co-branded cards in weeks and approve more than 90 per cent of applicants. According to the founders, this contrasts sharply with traditional Indonesian banks, which often require years to launch similar products and typically approve fewer than 5 per cent of applicants.

Honest draws inspiration from successful digital card models in the Americas, such as Nubank, Ramp, and Imprint, but tailors its approach specifically for the Asian market.

Ross Fubini, Managing Partner at XYZ Venture Capital, said: “Honest has solved problems traditional banks couldn’t touch, and you can see the difference in how people talk about the product—they love using it.”

Also Read: Navigating Southeast Asia’s digital economy in 2025: Trends, growth and innovation

Following the round, Fubini will join Honest’s board, with XYZ becoming the company’s second-largest investor, only trailing Orico. The firm’s other investors are Village Global, Goodwater, Rakuten, GMO, and David Vélez.

Recently, another Indonesian credit card startup Skor Technologies raised US$6.2 million in a pre-Series A funding round led by Argor Capital.

The post Alpha JWC co-founder’s credit card startup Honest secures US$100M appeared first on e27.

Despite weaker-than-expected private payroll data and the onset of a US federal government shutdown, risk appetite remained surprisingly resilient. This resilience is not born of complacency but rather of a recalibration in expectations around monetary policy, particularly the growing conviction that the Federal Reserve may soon pivot toward rate cuts.

The ADP National Employment Report showed a decline of 32,000 private-sector jobs in September, following a revised 3,000 decrease in August, standing in stark contrast to the median Bloomberg survey forecast of a 51,000 gain. This miss reinforced market bets that the labour market is cooling, thereby increasing the likelihood of a dovish shift from the Fed later this month.

The immediate market reaction was telling: US Treasury yields fell, with the 10-year yield dropping 5.2 basis points to close at 4.098 per cent, while the US Dollar Index edged down 0.07 per cent to 97.7. Simultaneously, gold surged to a record high of US$3,865.70 per ounce, a classic safe-haven move that also signals growing confidence in lower-for-longer rate expectations.

Equity markets responded with cautious optimism. Wall Street closed higher on Wednesday, with the Dow Jones gaining 0.09 per cent, the S&P 500 up 0.3 per cent, and the Nasdaq climbing 0.4 per cent. The healthcare sector provided strong support, suggesting investors are rotating into defensive yet growth-oriented segments amid macro crosscurrents.

Asian equities followed suit, mainly ending higher and continuing their upward trajectory in early Thursday trading, led by gains in semiconductor and broader technology stocks. US equity index futures pointed to further upside at the open, underscoring a broader narrative: markets are pricing in a soft landing scenario, where economic data deteriorates just enough to prompt Fed accommodation without triggering a full-blown recession.

This nuanced outlook has created fertile ground for alternative assets, particularly cryptocurrencies, which have begun to reassert their role not just as speculative instruments but as potential macro hedges.

Also Read: Diverging signals: Dow rises, gold breaks records, and crypto faces derivatives squeeze

The crypto market rose 3.91 per cent over the past 24 hours, extending a seven-day gain of 4.11 per cent. This sustained rally is not driven by retail FOMO alone but by structural developments that signal deeper institutional entrenchment and regulatory progress.

Three key catalysts stand out: the launch of institutional-grade Bitcoin options, regulatory maturation in Asia, particularly Hong Kong, and a surge in decentralised finance (DeFi) liquidity through major platform integrations. Each of these factors contributes to a more robust and credible ecosystem, one that increasingly appeals to traditional finance participants seeking exposure to digital assets without compromising on risk management or compliance.

The debut of Bitcoin options on Bullish Exchange on October 8 marks a significant milestone in the institutionalisation of crypto. Backed by heavyweight players such as BlackRock, Galaxy, Cumberland, and Wintermute, this offering arrives at a time when open interest in crypto derivatives has already reached a yearly high of US$1.24 trillion, up 30 per cent month-over-month.

Weekly inflows into Bitcoin ETFs reached US$571 million, further validating demand from regulated investment vehicles. Options markets deepen liquidity, enable sophisticated hedging strategies, and reduce volatility over time by allowing large players to manage risk without selling spot holdings.

The immediate market response was telling: perpetual funding rates surged 207 per cent within 24 hours, indicating a sharp increase in leveraged long positioning. This suggests that institutional participants are not just passively investing but actively expressing bullish macro views through derivatives. If trading volume on the new options platform proves robust, it could cement Bitcoin’s status as a legitimate macro hedge akin to gold but with asymmetric upside potential in a low-rate environment.

Also Read: The future of blockchain technology goes beyond just cryptocurrency and NFTs

Parallel to this institutional build-out, Asia is emerging as a critical regulatory laboratory for crypto adoption. Hong Kong’s Monetary Authority (HKMA) has received 36 applications for stablecoin licenses, with submissions coming from established banks and major tech firms.

This signals a shift from regulatory ambiguity to structured oversight, a prerequisite for large-scale institutional capital deployment. Stablecoins serve as the on-ramp and off-ramp for digital asset ecosystems, and their formal regulation removes a major friction point for traditional finance integration.

In South Korea, SK Planet’s adoption of Moca Network’s decentralised identity system triggered a 60 per cent rally in ZEN, illustrating how real-world utility can drive value in privacy-focused protocols. Crucially, crypto-equity correlations remain elevated at +0.76 against the Nasdaq, meaning that positive sentiment in tech equities continues to spill over into digital assets. As Asian regulators provide clearer guardrails, they reduce the jurisdictional risk that has long deterred pension funds, asset managers, and corporate treasuries from entering the space.

Meanwhile, DeFi is experiencing a quiet but significant expansion in accessibility. Coinbase’s integration of 1inch’s Swap API now grants its users access to millions of tokens across decentralised exchanges. This move contributed to a 17.92 per cent spike in spot trading volumes, though derivatives still dominate 84 per cent of total crypto volume.

The integration lowers the barrier to entry for retail investors seeking exposure to emerging narratives such as privacy coins like Zcash, which jumped 60 per cent. However, the Altcoin Season Index dipped 3.23 per cent, suggesting that while capital is exploring beyond Bitcoin and Ethereum, it has not yet committed to a broad-based rotation.

This hesitation may reflect lingering caution or simply the time lag between infrastructure development and narrative adoption. Either way, the trend points toward a more interconnected and liquid DeFi landscape, where centralised platforms act as bridges to decentralised liquidity.

Also Read: The Fed’s first rate cut: What it means for equities, risk, and crypto

Taken together, these developments paint a picture of a maturing asset class. The current rally is not a speculative bubble but a reflection of tangible progress on multiple fronts: institutional infrastructure, regulatory clarity, and technological interoperability. The confluence of Bullish Exchange’s options launch, Hong Kong’s stablecoin licensing momentum, and Coinbase’s DeFi integration represents a trifecta of credibility-building measures.

These are the foundations upon which a sustainable, long-term bull market can be built, not on hype, but on infrastructure. The path forward will not be linear, and leverage remains a double-edged sword, but the structural tailwinds are stronger than they have ever been. Traders must remain vigilant.

Open interest has risen 14 per cent in a single day, indicating that leverage is building rapidly. In a market still sensitive to macro surprises, a sudden shift in sentiment, perhaps triggered by stronger-than-expected US jobs data, could spark a short squeeze or a wave of liquidations.

The upcoming US nonfarm payrolls report, though potentially delayed due to the government shutdown, remains a critical inflection point. fA weak print would likely reinvigorate rate-cut expectations, further boosting risk assets and strengthening the correlation between crypto and traditional markets. Conversely, a resilient labor market could force a reassessment of the dovish narrative, testing the durability of this rally.

In essence, the crypto market is at a crossroads. It is no longer solely driven by retail enthusiasm or macro liquidity cycles. Instead, it is being reshaped by institutional architecture, regulatory milestones, and real-world utility. As such, the current price action should be viewed not as a fleeting surge but as the market pricing in a new phase of digital asset evolution.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. Share your opinion by submitting an article, video, podcast, or infographic.

Enjoyed this read? Don’t miss out on the next insight. Join our WhatsApp channel for real-time drops.

Image courtesy:

The post Record gold, falling yields, and rising Bitcoin: The interwoven narrative of modern risk assets appeared first on e27.

Singapore-based Neptune Robotics, which specialises in robotics-driven vessel cleaning, has raised US$52 million in Series B funding.

Granite Asia led the round. NYK Line, one of the world’s largest shipping companies, also participated.

The funding is earmarked to fuel research and development (R&D), the creation of new robotic systems, AI-driven service platforms, and expansion into 20 markets worldwide, with Japan identified as a potential key hub.

Also Read: Rise of the machines: 20 robotics startups shaping Southeast Asia’s future

Biofouling–the accumulation of marine organisms on a ship’s hull–is a massive economic and environmental burden. It increases fuel consumption by up to 30 per cent, costing the global shipping industry an estimated US$40-50 billion annually and driving excess emissions.

Neptune Robotics’s AI-powered underwater robots address this challenge by cleaning ship hulls 3-5 times faster than divers. The systems can clean full-draft capesize vessels in under 24 hours, operate day and night in clear or murky waters, and manage strong currents of up to 4 knots—a capability four times that of conventional diving.

Neptune helps clients cut fuel use, lower emissions, and advance their 2050 net-zero goals. It serves the world’s top bulk carrier and container fleets. Its systems operate across 61 Asian ports, covering Singapore and China, representing around 70 per cent of major trade routes.

According to Neptune Robotics, its robots can deliver up to 10x ROI, cut emissions, and remove humans from dangerous work.

Neptune’s partnership with NYK Line signals that robotics-enabled decarbonisation is ready to scale. NYK, which has worked with Neptune since 2022, is now rolling out the technology across its global fleet of bulk carriers, car carriers, and other carriers.

Also Read: Why robotics is just entering its prime phase

Elizabeth Chan, CEO of Neptune Robotics, commented on the investment: “Granite Asia and NYKs’ support validates how far robotics has come in transforming maritime efficiency and sustainability. This round gives us the resources to scale globally, continue innovating, and help shipowners boost returns while cutting emissions.”

The post Singapore’s Neptune Robotics secures US$52M to fuel global rollout of AI-powered vessel cleaning appeared first on e27.

Working at a small, mission-driven company has given me perspective on what I actually want from my career. There’s no free lunch, no ping pong tables, and no equity that might make me rich. But there’s also no bureaucracy slowing down good ideas, no projects that get cancelled after months of work, and no feeling that my contributions are just a drop in an ocean.

When I fix a bug or launch a feature, I can see its immediate impact. When someone successfully completes a purchase using our platform, or when a tool we built helps someone take the next step, I know my code played a part in that moment. That direct connection between effort and outcome has been incredibly motivating.

Constraints have also made me more resourceful. When you can’t just spin up another service or hire a specialist, you learn to find creative solutions. It’s pushed me to become a more thoughtful architect and a more efficient developer.

Also Read: Transition climate risk: Navigating the future of sustainable real estate

The post What a niche startup taught me about tech, growth, and impact appeared first on e27.