Rapid advances in Artificial Intelligence are likely to change the world for both good and ill.
Artificial Intelligence (AI) research is booming.
AIs now include a system's ability to:
- Adapt and improvise in a new environment.
- Generalise its knowledge.
- Apply knowledge to unfamiliar scenarios.
AI has intelligence characteristics include:
- Knowledge representation.
- Motion and manipulation.
Types of AI
- Narrow AI does defined tasks as demonstrated by Siri (a virtual assistant) using speech recognition.
- General AI, such as HAL in the movie 2001, uses human-style, adaptable intellect.
- Evolves as per Darwin's theory of natural selection.
- Solves problems using random mutations and combinations.
- Expert systems that use rules and a large number of inputs to make decisions, e.g. an autopilot.
Recent AI developments.
- AlphaGo Zero: has mastered chess and shogi.
- OpenAI: has AI agents that invented their own language.
- Facebook: has trained agents to negotiate and lie.
- Generative Pre-trained Transformer 3: generates articles on almost any topic.
- AlphaFold 2: rapidly examines amino acids to derive that protein's 3D structure for medical use.
- Google, Microsoft, and Tesla: are using specialised chips tailored to running and training machine-learning models.
- Google's TensorFlow: allows the public to build machine learning models.
- Development of Generative Adversarial Networks: means smaller amounts of labelled data (and hence fewer people) for machine-learning systems.
Key AI terms
- Machine learning: a computer system learns how to perform a task (such as predicting house prices) instead of being programmed.
- Neural networks: loosely based on the connections between neurons in the brain. The network learns how to carry out a particular task.
- Supervised learning: vast amounts of existing data is annotated by online workers (e.g. 50,000 through Amazon Mechanical Turk) to highlight the features of interest and then apply to new data.
- Unsupervised learning: algorithms try to identify patterns in data, looking for similarities that can be used to categorise that data. Google News uses this to group together stories.
- Reinforcement learning: trial and error is used until the best outcome is achieved.
Implications of AI
- Manual labour, checkout personnel and administrative/repetitive jobs are likely to be replaced.
- Essay writing, truck driving, and surgery will probably be undertaken by AI in the long term.
- According to Stephen Hawking, AI threatens our existence.
- The massive energy consumption of machine learning is contributing to global warming.
- Can reinforce gender and human racial biases.
- Abuse of facial recognition surveillance.
- Neural networks are creating fake news by mimicking an individual's voice and creating photo-realistic images.
- Healthcare is likely to become even more effective as AI can help to develop and test new drugs by spotting genetic sequences.
- Speaking computers will become commonplace.
- Chinese firms, Alibaba etc., compete with Google, Apple and Microsoft.
- Individuals can create their machine learning programs through training and infrastructure already available.
- May 2021 Edgy 8 Examples of Artificial Intelligence in our Everyday Lives
- Aug 2021 TDS 5 Must-Know AI Concepts In 2021
- Aug 2021 Analytics Insight Use of Artificial Intelligence and Machine Learning in NASA
- Aug 2021 The Conversation In a world first, South Africa grants patent to an artificial intelligence system
- Aug 2021 Nature AI drug discovery booms in China
Web Sites that provide articles relating to AI: