A number of jokes PaLM from Google has been able to explain and explain.

You may have missed five recent AI breakthroughs

The past two years have been a huge success for AI creators, as researchers and engineers have shown that AI can be applied to improve almost anything.

2021 saw a $93.5 billion injection of private equity into artificial intelligence, according to the Stanford University Artificial Intelligence Index 2022 report. Investment in artificial intelligence is expected to reach $500 billion by 2024, according to the International Data Corporation. PricewaterhouseCoopers estimates that the global artificial intelligence market will generate $15.7 trillion by 2030.

With years of intense investment by both public and private interests, AI is rapidly establishing itself in new areas including health, robotics, agriculture and many more.

While learning machines continue to improve on a regular basis, AI still has to meet the high expectations most of them have. While AI can excel in a specialized task, Artificial General Intelligence (AGI) is still a long way off, for now.

Artificial General Intelligence is the holy grail of every AI researcher, and is generally understood as an AI capable of understanding, learning, or performing any task that a human can do. However, in the real world, algorithms fall short when faced with complexity and the possibility of change.

Despite the limitations, companies like Google, Facebook, and Amazon have invested billions in this field, as 2022 saw major breakthroughs for artificial intelligence in multiple areas.

1. Artificial intelligence gets jokes

The single biggest advance in AI came after Google released an AI natural language processing model worth 540 billion variables that exceeds average performance. Larger models are more efficient at “transfer learning”, which seeks to train neural networks that use less data and computing power.

A number of jokes PaLM from Google has been able to explain and explain. (The Google)

This replaces OpenAI’s competitor GPT-3 natural language processing model with 175 billion machine learning parameters. GPT-3, originally introduced in May 2020, was hailed for being so accurate that it was difficult to tell if its text was written by a human. On challenging problems of common sense and reasoning.

2. AI becomes artistic

DALL E 2 is a new artificial intelligence system developed by OpenAI that can draw realistic portraits and arts based on a text description you provide. First introduced in 2021, the second iteration of the system produces creative images that are 4 times more accurate.

Image drawn by DALL E 2 based on the text:

Image drawn by DALL E 2 based on text: “An astronaut playing basketball with cats in space as a child illustration.” (Dalle 2 / OpenAI)

Created to depict how artificial intelligence sees the world and help people express creativity, DALL E 2 quickly raises the question of what it means to be human and creative.

This builds on previous advances in AI-assisted optical processing technology, including converting black and white photos to color, or creating lifelike 3D models of people from old photos. DALL E 2 is the latest “multimedia” system capable of working with both images and text.

3. AI can make streaming cheaper

Deepmind is the company that brought you AlphaGo, which defeated world champion Go Lee Sedol, before eventually leading to MuZero who mastered complex games like Chess, Shogi, Atari and even strategy games like Starcraft.

The same company also developed Artuµ, which will be used by the USAF as a spy plane’s radar operator, co-pilot and mission planner. MuZero recently addressed the challenge of video compression, reducing the amount of data required to stream video by 4 percent. This is not a small amount, given that standard compression codecs have been achieved after decades of engineering.

Analysts point out that video streaming made up the majority of Internet traffic in 2021. With only video streaming expected to grow in the coming years, more effective video compression can reduce streaming costs, and increase load power efficiency.

4. Competition for AI Autonomous Driving Intensifies

This year it will see a race from San Francisco to New York between self-driving car companies Tesla, Waymo and Cruz. In a showdown between competing methods, Tesla is set to remove its radar sensors in favor of a vision-based system.

A picture of how Tesla's fully autonomous driving AI (FSD) sees the world.

A picture of how Tesla’s fully autonomous driving AI (FSD) sees the world. (Tesla)

Radar can see through fog and snow, unlike cameras, indicating a deep confidence in Tesla’s AI. Other competitors have added more sensors, with both approaches set to test versus the other.

Waymo uses LiDAR technology, similar to radar, but uses laser pulses instead of radio waves. Tesla claims it’s easier to analyze sensor data, providing a multifocal view of the road beyond human reaction and vision. The company recently announced that its self-driving taxi trial in San Francisco was a success.

Cautious optimism

As AI systems slowly move away from specialization to hybrid systems, the unresolved challenge most AI researchers face is finding a way to integrate knowledge from multiple sources into consistent outputs.

Models of language used to interact with humans, answer questions, and write persuasively are being prompted quickly by major technology investments in voice assistants like Alexa and Siri.

While 2022 is far from over, there is a clear trend towards bringing together the traditional challenges of speech, cognition and language rather than treating them as separate functions. This is harder than it sounds, but if the past few years are any indication, AI will continue to drive innovation in multiple sectors for decades to come.

Source: TRT World

#missed #breakthroughs

Leave a Comment

Your email address will not be published.