Deepfakes Technolog Topic 1

How Deepfakes work

Deepfake content is created by using two competing AI algorithms -- one is called the generator and the other is called the discriminator. The generator, which creates the phony multimedia content, asks the discriminator to determine whether the content is real or artificial.

Together, the generator and discriminator form something called a generative adversarial network. Each time the discriminator accurately identifies content as being fabricated, it provides the generator with valuable information about how to improve the next deepfake.

The first step in establishing a GAN is to identify the desired output and create a training dataset for the generator. Once the generator begins creating an acceptable level of output, video clips can be fed to the discriminator.

As the generator gets better at creating fake video clips, the discriminator gets better at spotting them. Conversely, as the discriminator gets better at spotting fake videos, the generator gets better at creating them. 

Until recently, video content has been more difficult to alter in any substantial way. Because deepfakes are created through AI, however, they don't require the considerable skill that it would take to create a realistic video otherwise. Unfortunately, this means that just about anyone can create a deepfake to promote their chosen agenda. For example, a deepfake could be used to spread false information via a presidential candidate. Microsoft, however, has worked on an AI-powered deepfake detection software for this purpose. The tool can automatically analyze videos and photos to provide a confidence score that the media has been manipulated.

Another possible danger deepfakes introduce is that people will take such videos at face value, and after realizing it’s fake,  people will stop trusting in the validity of any video content at all.

What is artificial general intelligence (AGI)?

Artificial general intelligence (AGI) is the representation of generalized human cognitive abilities in software so that, faced with an unfamiliar task, the AGI system could find a solution. The intention of an AGI system is to perform any task that a human being is capable of.

Definitions of AGI vary because experts from different fields define human intelligence from different perspectives. Computer scientists often define human intelligence in terms of being able to achieve goals. Psychologists, on the other hand, often define general intelligence in terms of adaptability or survival.

AGI is considered to be strong artificial intelligence (AI). Strong AI contrasts with weak or narrow AI which is the application of artificial intelligence to specific tasks or problems. and self-driving cars are examples of narrow artificial intelligence.

What can artificial general intelligence do?

AGI in computer science is an intelligent system with comprehensive or complete knowledge and cognitive computing capabilities. As of right now, no true AGI systems exist; they remain the stuff of science fiction. The performance of these systems is indistinguishable from that of a human, at least in those terms. However, the broad intellectual capacities of AGI would exceed human capacities because of its ability to access and process huge data sets at incredible speeds.

True AGI should be capable of executing human-level tasks and abilities that no existing computer can achieve. Today, AI can perform many tasks but not at the level of success that would categorize them as human or general intelligence.

An AGI system should have the following abilities:

• abstract thinking
• background knowledge
• common sense
• cause and effect
• Transfer learning

• Practical examples of AGI capabilities include the following five:

1. Creativity. An AGI system would theoretically be able to read and comprehend human-generated code and improve it.
2. Sensory perception. AGI would excel at color recognition, which is a subjective kind of perception. It would also be able to perceive depth and three dimensions in static images.
3. Fine motor skills. An example of this includes grabbing a set of keys from a pocket, which involves a level of imaginative perception.
4. Natural language understanding (NLU). Understanding human language is highly context-dependent. AGI systems would possess a level of intuition that would enable NLU .
5. Navigation. The existing Global Positioning System (GPS) can pinpoint a geographic location. Once fully developed, AGI would be able to project movement through physical spaces better than existing systems.

AI researchers also anticipate that AGI systems will possess higher-level capabilities, such as being able to do the following:

handle various types of learning and learning  algorithms;

create fixed structures for all tasks

understand symbol systems;

use different kinds of knowledge;

understand belief systems; and

engage in metacognition and make use of metacognitive knowledge

AGI vs. AI: What's the difference?

AGI should theoretically be able to perform any task that a human can and exhibit a range of intelligence in different areas. Its performance should be as good as or better than humans at solving problems in most areas of intelligence.

In contrast, weak AI excels at completing specific tasks or types of problems. Many existing AI systems use a combination of machine learning, deep learning, reinforcement learning, and natural language processing for self-improving and solving specific types of problems. However, these technologies do not approach the cumulative ability of the human brain.

• customer service chatbots ;
• voice assistants  like Siri and Alexa;
• recommendation engines such as those Google, Netflix, and Spotify use;
• marketing platforms used to gather business intelligence  and customer sentiment ; 
• facial recognition applications.

Examples of artificial general intelligence

True AGI systems are not on the market yet. However, examples exist of narrow artificial intelligence systems that approximate or even exceed human abilities in certain areas. Artificial intelligence research is focused on these systems and what might be possible with AGI in the future.

Here are some examples of that system.

• IBM's Watson. Watson and other supercomputers are capable of calculations that the average computer can't handle. They combine their immense computing power with AI to carry out previously impossible science and engineering tasks, such as modeling the Big Bang theory of the birth of the universe or the human brain.
• Expert systems. These systems are AI-based ones that mimic human judgement. They can recommend medicine based on patient data and predict molecular structure.
• Self-driving cars. These are able to recognize other vehicles, people, and objects in the road and adhere to driving rules and regulations

• ROSS Intelligence. ROSS is a legal expert system that is also called the "Al attorney." It can mine data from about 1 billion text documents, analyze the information and provide precise responses to complicated questions in less than three seconds.
• AlphaGo. This is another example of narrow intelligence that excels at a specific type of problem-solving.AlphaGo is a computer program that can play the board game Go. Go is a complex game that is difficult for humans to master. In 2016, AlphaGo beat the world champion Lee Sedol in a five-game match.
• Language model Generative Pre-trained Transformer 3. GPT-3  is a program that can automatically generate human language. In some cases, the text is indistinguishable from the human output, but the output is often flawed. The technology is consistently able to emulate general human intelligence.
• Music AIs. Dadabots is an AI algorithm that, given a body of existing music, can generate a stream of its own approximation of that music.

If AGI was applied to some of the preceding examples, it could improve their functionality. For example, self-driving cars require a human to be present to handle decision-making in ambiguous situations. The same is true for music-making algorithms, language models, and legal systems. These areas include tasks that AI can automate but also ones that require a higher level of abstraction and human intelligence.

What is the future of AGI?

Many experts are skeptical that AGI will ever be possible. Others question whether it is even desirable.

English theoretical physicist, cosmologist, and author Stephen Hawking warned of the dangers in a 2014 interview with the British Broadcasting Corp. "The development of full artificial intelligence could spell the end of the human race," he said. "It would take off on its own and redesign itself at an ever-increasing rate. Humans, who are limited by slow biological evolution, couldn't compete and would be superseded."

However, some AI experts expect the continued development of AGI. In an interview at the 2017 South by Southwest Conference, inventor and futurist Ray Kurzweil predicted computers will achieve human levels of intelligence by 2029.

The Church-Turing thesis, developed by Alan Turing and Alonzo Church in 1936, is another perspective that supports the eventual development of AGI. It states that, given an infinite amount of time and memory, any problem can be solved using an algorithm. Which cognitive science algorithm will be is up for debate. Some say neural networks show the most promise, while others believe in a combination of neural networks and rule-based systems

Another potential initiative comes from neuroscience: neuromorphic computing, which uses artificial neurons and synapses to replicate the biological framework and functioning of the human brain.