Yes, but only in a qualified sense. An AI can discover strategies, algorithms, or hypotheses that were not supplied as human examples when it has a defined environment, rules or observations, an objective, and feedback. AlphaGo Zero learned Go entirely through self-play and defeated an earlier human-trained AlphaGo program 100–0. That is genuine novelty relative to human demonstrations. It is not evidence that an AI can produce reliable, unconstrained scientific knowledge from nothing: open-ended discovery still depends on data, experiments, formal rules, or a human-selected goal, and people must verify that a result is both correct and historically new.
What “genuinely new” can mean
The phrase combines several different tests. Confusing them leads either to exaggerated claims that AI “invented knowledge” or to an overly narrow view that nothing is new unless a person first imagined it.
As an Amazon Associate I earn from qualifying purchases.
Novel to the system
A result is novel to the system when it has not encountered that move, design, hypothesis, or rule in its own training trajectory. This is the weakest form of novelty. A reinforcement-learning agent may rediscover a known tactic without being shown it directly.
Novel relative to human examples
A result meets a stronger test when the system did not learn by copying demonstrations, labels, or expert solutions. AlphaGo Zero satisfies this standard within Go: it received the board position, legal-move rules, and the win-or-loss outcome of games it generated itself, rather than records of human play.
#1 Best Overall
- [Ultra-Lightweight Ear Buds Designed for Small Ears] Each earbud weighs only 3.7g and features a compact, ergonomic in-ear design made especially for small ears. Secure, low-profile, and comfortable for workouts, all-day wear or overnight sleeping.The ultra-slim profile sits flush in your ear with no hard edges poking or rubbing, so you can even sleep on your side without discomfort.
- [Immersive Stereo Sound with TOZO OrigX Technology] TOZO OrigX tuning delivers clear vocals, balanced mids, and natural stereo sound for music, podcasts, and videos.
- [Long Battery Life for Daily Use] Get up to 7 hours of playtime on a single charge, with up to 32 hours total using the charging case—ideal for workdays, commuting, and extended listening sessions.
- [Bluetooth 5.3 & Stable Connection] Bluetooth 5.3 provides fast pairing, stable wireless performance, and reduced dropouts as you move around home or office.
- [Deep Bass with Clear Vocals] High-performance drivers produce punchy bass while keeping vocals clean and detailed for everyday listening.
Historically or scientifically unprecedented
The strongest claim is that no person had previously conceived the result and that it remains valid after independent checking. Current evidence does not provide a universal test for proving this. A system can produce an unfamiliar strategy while humans later discover that an old paper already described it, or that the strategy fails outside the benchmark.
Therefore, “new” should always be qualified: new to the model, new compared with supplied examples, or new to the world after independent validation.
How discovery without human examples works
Removing demonstrations does not remove all information. It changes where the learning signal comes from.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Self-play and reinforcement learning
In self-play, an agent repeatedly acts in an environment, observes the consequences, and updates its policy to increase a numerical reward. In a game, the environment supplies legal moves and the final result supplies the reward. The agent can therefore generate its own training data.
- Define the state and actions. The program receives a formal representation, such as a Go board and the legal moves.
- Choose an objective. Winning, maximizing a score, or satisfying a constraint becomes the reward signal.
- Explore. The agent tries many actions, including moves that human experts would normally reject.
- Evaluate outcomes. Games, simulations, or experiments provide feedback about which choices helped.
- Update the policy. The system increases the probability of actions associated with better results.
- Repeat at scale. Millions of trials let the agent search combinations that are impractical for a person to enumerate.
The resulting behavior can be unfamiliar because no human move list limits the search. Yet the search remains bounded by the representation, rules, reward, computing budget, and stopping criteria chosen in advance.
Rank #2
- 【IP68 Waterproof Multifunctinal Smart Watch】Torjalph fitness activity tracker smart watch is IP68 waterproof technology ,you can wear it when swimming,handwashing,it's rain-proof which means that you perform multiple exercise modes outside without worrying about sweat or rain corroding your fitness smart watch,and this watch equipped with a 230 mAh large capacity battery magnetic charging ,and this sports activity tracker watch can be used 5-7 days If fully charged.
- 【Larger 1.83'' Full Screen Watch & Multi-dial Selection】1.83 inch full HD color touch screen smart watch for men / women provides excellent 240*284 resolution image quality ,This Bluetooth smart watch offers a better touch experience and remarkable immersion and ,more than 200 watch faces for you to choose from App"Da Fit ", you can change the dial style according to your mood to wear keeping the world on your wrist
- 【Accurate Activity Tracker Smart Watch 】sports watch activity tracker uses sweat and dust resistant breathable silicone strap, our Fitness Tracker Smart Watch for Android iOS Phones supports 12 sports modes,walking,running,bicycle,swimming,badminton,basketball,football,rope skipping. This bluetooth calls smart watch for men women is lightweight and stylish, bringing more convenient and relaxing sports experience while also providing more protection.
- 【Your 24 Hour Personal Daily Health Monitor】Fitness watch for women men full touch screen IP67 waterproof bluetooth smartwatch for Android iOS Phones with heart rate blood pressure monitor multifunction Andriod smart watch with steps counter sleep monitor sports fitness watch, monitoring blood oxygen saturation allows you to understand your physical condition in real time to track fitness progress ,to know your health status in time.
- 【More Functions Bluetooth Call Smart Watch】This fitness activity smart watch full screen for iPhone and Android phones includes a bunch of new features including:The bluetooth smart watch can hang up the phone, receive and display WhatsAPP /Text /Facebook /Twitter /Messenger reminder.Support sedentary reminder, alarm clock, stopwatch, timer,weather information,music control, remote camera control compatible with Android 5.0 and above or iOS 9.0 and above smartphones.
Search guided by populations of agents
The same principle can operate at the level of learning algorithms rather than game moves. A 2025 Nature study reported a reinforcement-learning update rule discovered from the cumulative experience of populations of agents. The discovered rule surpassed established update rules on Atari and other benchmarks. The agents were not handed a human-designed formula to imitate; they searched a space of possible updates and retained those that improved measured performance.
This is algorithmic novelty, not unconstrained intelligence. The benchmark environments, scoring functions, and comparison rules defined what “better” meant. A rule that wins on Atari may not transfer to robotics, finance, or scientific modeling without new evidence.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
AlphaGo Zero: a clear case of novelty without human demonstrations
DeepMind’s AlphaGo Zero is the most direct demonstration available in the evidence summarized here. The 2017 Nature study describes an algorithm based solely on reinforcement learning, “without human data, guidance or domain knowledge beyond game rules.” It started from random play, used self-play to improve, and reached superhuman strength.
What it achieved
- AlphaGo Zero defeated the earlier, human-trained AlphaGo by 100–0 in the reported match.
- It developed strategies and moves not supplied in human game records.
- DeepMind reported that the probability of AlphaGo’s famous Move 37 in the Lee Sedol match was about 1 in 10,000 according to its estimate, illustrating how an unusual move can still be strategically strong.
- Go has approximately 10170 possible board configurations, making exhaustive human enumeration impossible.
DeepMind characterized the result as discovering “unconventional strategies and creative new moves.” Within the closed rules of Go, that is meaningful discovery: the system searched trajectories no human supplied and found lines of play that experts initially considered strange.
What AlphaGo Zero did not prove
The program did not discover the rules of Go, invent the concept of winning, or decide which board game mattered. Humans supplied the state representation, legal-move rules, reward definition, neural-network architecture, and evaluation procedure. Its novelty is genuine relative to human examples, but it is not discovery from an information-free void.
Rank #3
- [Low Noise High Fidelity] New bee guitar cord with 22AWG oxygen-free-copper center conductors provides enhanced signal transfer, the PVC insulation material and the high-density stranded copper double shield provide noise rejection for quiet operation
- [Ideal for Choice] New bee 1/4 straight and right angle metal connectors, ensures a secure connection in any instrument, pedal, or amp jack. L shape great fro on stage guitar stands
- [Wide Compatible] Suitable for mixer, power amplifier, microphone, effector, decoder, equalizer, electoronic organ, electric guitar, electric bass and other electronic equipments. Widely used in the installation of smart home, karaoke speaker, professional stage audio, etc
- [Easy Disassemble] Making incidental repairs quick and easy! Very nice, flexible cable
- [Dont Hesite to Order] 24 hours customer services and professional technology team are standing by
Games versus science: why the boundary matters
Games provide unusually favorable conditions for autonomous discovery. The state is precisely defined, actions are usually observable, outcomes arrive quickly, and a single score gives an unambiguous ranking. Scientific reality is noisy, partially observed, expensive to probe, and full of competing explanations.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsWhat a scientific system can receive
An AI looking for a physical law may be given measurements, a simulator, laboratory results, or a set of equations. It may then search for compact relationships, estimate parameters, or propose conservation laws. A 2020 Physical Review Letters demonstration showed that machine-learning systems can recover useful parameters and conservation laws from data. Such studies are important, but published examples remain constrained and often use toy problems rather than unrestricted laboratory discovery.
Why a plausible equation is not a discovery
A candidate law must survive tests that a game move does not face:
- Measurement quality: sensor noise, missing values, calibration errors, and selection bias can create spurious patterns.
- Generalization: the relationship must predict observations outside the data used to fit it.
- Replication: independent teams should obtain the same result with independent instruments or datasets.
- Comparison with prior work: specialists must check whether the claim is already known under different notation.
- Mechanistic coherence: the proposed explanation should fit established constraints or clearly show where they fail.
- Experimental confirmation: when feasible, a new experiment must distinguish the hypothesis from competing explanations.
A system can therefore be creative and useful before it is entitled to the label “scientific discovery.” The label belongs to the validated claim, not merely to the novelty of the generated text or equation.
Three systems compared
| System or approach | Learning signal | Structure supplied in advance | How success is validated | What kind of novelty is demonstrated |
|---|---|---|---|---|
| Human-example imitation | Demonstrations, labels, or expert solutions | Task definition plus curated examples | Accuracy against held-out examples or human ratings | Usually variation or recombination of known examples |
| AlphaGo Zero-style self-play | Game outcomes generated by the agent | Formal rules, board representation, and win objective | Strength against established players or programs | New strategies relative to human demonstrations in a closed task |
| Population search for an RL update rule | Cumulative performance of agent populations | Benchmark environments, candidate-rule space, and score | Performance on Atari and other reported benchmarks | A rule not hand-supplied by researchers that outperforms baselines on those tests |
| AI-assisted physical-law discovery | Measurements, simulations, or experimental feedback | Variables, model class, priors, and researcher-selected objective | Prediction, replication, and independent experiments | Potentially new relationships, but demonstrations remain constrained and validation-intensive |
Does AI need human data to discover something?
No. It needs information, but that information can come from interaction rather than a human example set. In Go, the rules and the consequences of moves contain enough structure for self-play. In a physics problem, measurements or a simulator can replace demonstrations. In a laboratory, the feedback may be the result of an experiment.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Rank #4
- Digital Family Calendar: BSIMB digital calendar wall touch screen puts an end to your chaotic life. Designed for busy families, the digital calendar keeps you organised and in step with your family. Assign different colours and tasks to each family member to make family scheduling easier and more intuitive.
- Easy Setup and Auto-Sync: Plug in, connect to WLAN and auto-syn your electric calendar in minutes. Digital wall calendar compatible with Google calendar, iCloud calendar, Outlook calendar, Cozi calendar and Yahoo calendar. Add events directly on the device or through the ecalendar app. Electronic calendar stays current whether you're at work or traveling—no manual updates needed.
- All-In-One Home Assistant: Smart calendar builds habits and encourage independence with personalised chores. Display daily meal plan to reduce the daily ‘What's for dinner?’ stress. To-do lists keep track of everything you want. Turning chores into valuable achievements with stars and rewards, making life easier and more enjoyable.
- Manage Anytime Anywhere: BSIMB electronic calendar for family stay connected with the powerful ecalendar app. Access and manage your digital wall calendar anytime, anywhere with free features like stars & rewards, photo screensaver, chore charts, meals plans, to-do lists etc. Electronic planner paid features like Magic Import tailor advanced tools for users with special needs.
- Stay in Touch with Loved Ones: Share photos or videos from your phone to electronic wall calendar with app, instantly update your life moment, and keep in touch with your family, especially those who are far away; you can also surprise your smart calendar digital wall by sending him a wish card; it's also great for keeping tech-illiterate elders in touch with the youngsters.
However, “no human examples” is not the same as “no human assumptions.” People still choose the environment, sensors, variables, reward, optimization method, safety limits, and criteria for stopping. Those choices determine which discoveries are reachable and which are invisible to the system.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to evaluate a claim that an AI discovered something new
Use the following procedure before accepting a headline claim.
- Specify the novelty baseline. Ask whether the result is new to the model, absent from its demonstrations, or unprecedented in the historical literature.
- Inspect the information boundary. Record every rule, dataset, simulator, pretrained component, prompt, reward, and human intervention available to the system.
- Separate generation from selection. A model may generate thousands of candidates while humans select the one worth reporting. State who performed each step.
- Use a holdout. Test the strategy, algorithm, or hypothesis on cases not used during search. For scientific work, reserve independent measurements or experiments.
- Run ablations. Remove self-play, human initialization, a data source, or a reward component to determine what actually caused the result.
- Check reproducibility. Publish the configuration, random seeds where appropriate, data-processing decisions, and evaluation code so another team can repeat the result.
- Search prior knowledge. Compare against papers, patents, databases, and expert review. A result can be independently rediscovered without being historically unprecedented.
- Test robustness and costs. Look for sensitivity to noise, distribution shifts, adversarial cases, compute budget, and measurement expense.
This process distinguishes a surprising output from a reliable contribution. It also makes clear which part of the claim is supported: novelty, performance, correctness, or historical priority.
Common failure modes
Calling random novelty a discovery
Unfamiliarity is not evidence of value. A move, formula, or design may be new only because it is bad. Discovery requires a measurable improvement, explanatory power, or a testable prediction.
Free tools Windows power users keep installed
One-click scans. No signup required.
Assuming a benchmark equals reality
Outperforming a baseline on a fixed benchmark demonstrates performance under that benchmark’s conditions. It does not establish broad superiority or transfer to an untested domain.
Ignoring hidden human examples
Pretraining data, engineered features, demonstrations embedded in prompts, and hand-designed reward functions can all leak human knowledge into a system. Claims of example-free learning should disclose these channels.
Confusing compression with explanation
An AI may fit a short mathematical expression to data without identifying the causal mechanism. Compactness can be useful, but it is not proof that the expression describes nature.
Overlooking the literature
Researchers often describe an output as new before checking older terminology, equivalent equations, or results in another field. Historical novelty requires a careful literature comparison, not just a new string of symbols.
Recommended Free Tools
Is AI creativity real?
Creativity is a useful description when a system explores combinations, strategies, or abstractions that its operators did not specify and that prove valuable under an objective. AlphaGo Zero’s unconventional moves meet that practical definition. The system did not copy a human game record, yet its play changed expert understanding of strong Go strategy.
That does not settle philosophical questions about imagination or consciousness. The operational question is narrower: can an AI produce an effective, previously unsupplied solution? In bounded environments, the evidence says yes. Whether the same process yields genuinely new, correct knowledge about the open world remains an empirical question requiring stronger validation.
What would count as open-ended autonomous discovery?
A convincing demonstration would need more than an impressive output. It would define a broad but realistic domain, allow the system to choose informative experiments, prevent access to the relevant human answers, require predictions on unseen conditions, and obtain independent confirmation. The record would also need to document the system’s data, objectives, interventions, and computational limits.
No reviewed evidence establishes a general-purpose AI that discovers open-ended scientific knowledge with no human-provided data, task definition, or evaluation signal. The demonstrated frontier is more precise and more useful: AI can search vast spaces, generate hypotheses, and discover strategies beyond human examples when the surrounding problem supplies structure and feedback.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
