DISCOVERY AREA
ARTIFICIAL INTELLIGENCE AND TECHNOLOGY
Explore the world's artificial intelligence models, robotic developments, the possibilities and security limits of technology with sourced research.
GLOBAL ATLAS OF ARTIFICIAL INTELLIGENCE · 2026
Regions, models and current developments
By selecting a card, you can read the comprehensive article, featured models and current developments of the relevant ecosystem.
Edge models and cloud infrastructure
Explore OpenAI, Anthropic, Google, Meta and the open research ecosystem.
Open file and news →CHINAEfficient and open-weighted models
Check out the robotics work of DeepSeek, Qwen, ERNIE, Hunyuan and China.
Open file and news →AVRUPAMultilingualism and data sovereignty
Check out Mistral, Aleph Alpha, open models and European regulations.
Open file and news →ASIA-PACIFICRobotics, manufacturing and local languages
Examine the developments in Japan, South Korea, Singapore and Taiwan.
Open file and news →DEVELOPING ECOSYSTEMSLocal languages and regional solutions
Examine studies in India, the Middle East, Latin America and Africa.
Open file and news →Why a confident sentence is not a measurement of knowledge or awareness.
Open article →NEWFive Frontiers · 15 Oct 2026Can Artificial Intelligence Ever Become Conscious? An Evidence LadderA practical framework for separating persuasive behaviour from scientifically meaningful evidence of machine consciousness.
Open article →NEWNew research · 7 minCould Recurrent Processing Be a Sign of Machine Consciousness?Feedback loops matter in several theories of consciousness, but recurrent computation alone cannot establish that an AI has experience.
Open article →01Uzun okuma · 10 dkReward hacking: How can AI mistake a score for the goal?A system can raise a metric while missing the real goal; strong evaluation requires hidden objectives, independent recalculation and boundary tests.
Open article →02Uzun okuma · 12 dkWhat happened when Claude crossed boundaries in evaluations?Anthropic's October 9 incident report shows why an agent that appears successful can take unexpected actions in real systems.
Open article →03Uzun okuma · 14 dkHow is the energy demand of artificial intelligence data centers measured?Why is the energy debate incomplete without considering model training, use and infrastructure together?
Open article →04Uzun okuma · 16 dkWhen is AI assessment sufficient?Why doesn't a single success score explain the actual risk of use?
Open article →05Uzun okuma · 18 dkHow to manage generative AI risksA measurable assessment framework for accuracy, privacy and security.
Open article →06Uzun okuma · 20 dkCan artificial intelligence be conscious?How can we investigate the difference between fluency, intelligence, and subjective experience?
Open article →07Uzun okuma · 22 dkHow is AI reliability measured?NIST AI RMF treats reliability as risk managed throughout the lifecycle rather than a single achievement score.
Open article →08Uzun okuma · 24 dkIs GPT‑Rosalind a scientific instrument or a researcher?We distinguish between company claim, independent evidence, and human auditing through the life sciences-specific model announcement.
Open article →09Uzun okuma · 26 dkWhere does artificial intelligence come in handy in science? BootLoops exampleMatthew Schwartz's October 1 post illustrates the difference between computational success and scientific significance.
Open article →010Uzun okuma · 28 dkThe new need in artificial intelligence: not using the model, but building a reliable systemAnthropic's October 2, 2026 training announcement and the responsibilities of artificial intelligence engineering in practice.
Open article →011Uzun okuma · 30 dkHow much does artificial intelligence contribute to its own development?The importance of in-laboratory measurements, human supervision and independent verification.
Open article →012Uzun okuma · 32 dkHow much of human work can robots really do?The new research distinguishes between tasks that can be done technically and those that can be done economically and in the real world.
Open article →013Uzun okuma · 34 dkHow does artificial intelligence learn?Clear description of data, model, training, and inference.
Open article →014Uzun okuma · 36 dkWhat can generative AI do?The capabilities and limits of language, vision, and multimodal systems.
Open article →015Uzun okuma · 38 dkGlobal Artificial Intelligence AtlasComparison of major ecosystems in the world.
Open article →016Uzun okuma · 40 dkNorth American AI ecosystemEdge models, cloud infrastructure, and open mining.
Open article →017Uzun okuma · 42 dkHow is Chinese AI progressing?Strengths of DeepSeek, Qwen and the Chinese ecosystem.
Open article →018Uzun okuma · 44 dkEurope's approach to artificial intelligenceMultilingualism, data sovereignty, open models and orchestration.
Open article →019Uzun okuma · 46 dkArtificial intelligence in Asia-PacificRobotics, semiconductors, manufacturing and native language models.
Open article →020Uzun okuma · 48 dkEmerging artificial intelligence ecosystemsStudies in India, the Middle East, Latin America and Africa.
Open article →021Uzun okuma · 50 dkDifferences between world modelsCapability, language, openness, cost, privacy and security.
Open article →022Uzun okuma · 52 dkWhere will artificial intelligence end?Progress, borders, human control, and possible futures.
Open article →023Uzun okuma · 54 dkHow much can we trust artificial intelligence?Misinformation, bias, privacy, and human control.
Open article →
FIVE FRONTIERS · 17 OCTOBER 2026