In just the last several years, artificial intelligence has completely revolutionized the world we live in. Cars can drive themselves, chatbots are having realistic conversations with humans, and computer-generated art is winning awards.
Every day, companies are figuring out new ways to try to capture the complexities of human intelligence in artificial systems, and then take them even further than that. It hasn’t quite been accomplished yet, but experts say it is only a matter of time before we achieve what is widely known as artificial general intelligence, or AGI — a form of artificial intelligence that is capable of processing information and performing actions in the same ways human beings can.
Artificial General Intelligence Companies to Know
- Google Brain
- Hanson Robotics
Artificial general intelligence can be approached and conceptualized in many different ways, but it boils down to creating a machine that can not only do what it is created to do, but also learn and carry out tasks outside of those parameters. These systems will have the ability to implement autonomous self control.
What Can Artificial General Intelligence Companies Do?
What does this look like in practice? Well, to understand what artificial general intelligence can do, one must first understand what AI is capable of at this moment.
Right now, all of the artificial intelligence out there — chatbots, self-driving cars, smart assistants — is considered to be weak, or narrow, AI. But don’t let the name fool you, this technology is anything but weak. These systems can beat world-champion chess players, finish typing our sentences for us, and prevent cyberattacks before they even happen. And they do all of this with the help of machine learning, deep learning and natural language processing.
But these systems are only capable of performing the designated tasks they were programmed to do. Granted, they may perfect that task (and even perform it better than any human ever could), but they only behave as desired under very specific, pre-defined conditions. If those conditions are altered in context, the model loses track of its purpose, which can lead to errors and, inevitably, human intervention. This limited functionality allows narrow AI systems to automate and perfect a specific task with ease, but it doesn’t allow for any other task to be executed with the same perfection.
As such, this does not accomplish the overall goal of AI, which is to replicate and simulate human intelligence in machines. After all, humans think dynamically and can adapt. We can solve more than one problem at a time and learn new skills along the way. And we can generalize knowledge, apply that knowledge from one task to another, plan ahead according to our current knowledge and adapt as changes occur.
The artificially intelligent systems we have in place today can’t do this — yet. But there are lots of companies out there working to make it a reality, and many industry leaders expect the emergence of AGI to be the obvious next step. Tech billionaire Elon Musk recently predicted we could achieve artificial general intelligence as soon as 2029, but many experts in both academia and the private sector think we are further out than that (some say it won’t occur until after 2060). Either way, while it is important to remember all the complex tasks AGI likely won’t be able to accomplish within the decade, it is undeniable that it stands to revolutionize virtually every industry if and when it comes to fruition.
Until then, these tech companies will continue to push the limits of artificial intelligence. Some are longtime innovators, others are fairly new on the scene. But all of them envision a world in which machines are just as intelligent and complex as humans are.
Artificial General Intelligence Companies to Know
As an AI safety and research company, Anthropic is dedicated to building reliable, understandable and maneuverable general AI. Its research interests span a variety of fields, including human feedback, reinforcement learning, natural language processing and code generation. For instance, one of Anthropic’s papers explored how to train a general language assistant to be helpful to users, without providing harmful advice or exhibiting bad behaviors — something certain AI systems have been wont to do at times. The company was founded in early 2022 by a former executive at OpenAI, another innovator in artificial general intelligence, and has already garnered quite a bit of financial backing.
Darktrace’s self-learning AI helps protect companies’ data and infrastructures from cyber threats by detecting them in real time. The platform works by analyzing network data and creating probability-based calculations, detecting deviations from typical behavior to identify threats. When Dartrace detects suspicious activity, it can put a stop to it before it causes any damage. Several major companies across the financial, healthcare, media and education sectors rely on Darktrace for protection, including Jimmy Choo, the Coca-Cola Company and Billabong.
DeepMind is perhaps best known for its creation MuZero, a computer program that uses AI to master games it has not even been taught to play through sheer brute force, re-playing games millions of times. But the company says it is working to “benefit humanity” with its artificial intelligence technology as well. Since its merger with Google in 2014, Deepmind has made advancements in medical research specifically, particularly as it relates to eye diseases.
Evolv Technology’s weapons detection scanner is designed to keep public venues safe. The portable system is able to screen hundreds of people an hour, allowing them to walk straight through at a normal pace, without stopping or having to remove anything from their pockets. Each machine is equipped with artificial intelligence and advanced computer vision, which is capable of detecting a wide range of metallic and non-metallic weapons. All the data collected from a network of sensors is processed on one software platform that is “constantly learning” according to the company, meaning it can adapt and become more intelligent as new threats are discovered. Over the years, Evolv’s technology has been implemented in schools, theaters, stadiums and other event spaces across the country, and has garnered support from several familiar names, including Bill Gates and a VC firm run by former Florida Governor Jeb Bush.
Graphcore has created a completely new processor — an intelligence processing unit, or IPU, which the company says will be an important part of the next step in AI’s evolution. The IPU is built in such a way that it speeds up AI computing, allowing software architects working in machine learning to undertake all kinds of projects without having to worry about the associated compute power and bandwidth. In the coming years, Graphcore expects its technology to drive innovation in a variety of industries that rely on AI, including finance, biotech and scientific research.
Working under the umbrella of the Google AI research division, Google Brain is a deep learning artificial intelligence research team. With about a half-dozen areas of interest, the team combines open-ended machine learning research with information systems and large-scale computing power to push the boundaries of what’s possible in the world of AI and ML. One of the team’s most successful projects to date was when it created one of the largest neural networks for machine learning ever, which eventually not only taught itself how to recognize cats, but was actually able to generate its own digital image of one — demonstrating that software-based neural networks could mirror human intelligence. Google Brain’s technology currently powers many products, including Google Translate, the search feature in Google Photos, video recommendations in YouTube and Android’s speech recognition system. And the group is continuing to forge new paths with AI in fields like healthcare, robotics and cryptography.
Hanson Robotics is at the forefront of AI and robotics, and aims to create socially intelligent machines that have “rich personalities and social cognitive intelligence” so they can “potentially connect deeply and meaningfully with humans,” according to its website. The company is working to accomplish this by developing cognitive architecture and AI-based tools that enable robots to simulate human personalities, have meaningful interactions with humans and evolve from those interactions. Hanson is perhaps best known for its creation of humanoid robot Sophia, which took the world by storm in 2016. Sophia has since provoked discussions about AI ethics and even inspired the creation of SophiaDAO, a decentralized autonomous organization that is intended to provide guidance for future artificial general intelligence development.
Powered by machine learning, Hyperscience automates office work. Essentially, its AI-base software extracts information from documents, turning human-readable content into machine-readable data so any given task, from data entry to client onboarding, can be done autonomously, without the need for human intervention. And Hyperscience says its technology continuously learns and evolves. So, as human involvement decreases, the software gets faster and smarter.
IBM was one of the first companies to really make headlines for its innovations in artificial intelligence. In 2011, its question-answering computer system named Watson was able to win a game of Jeopardy! against two former champions using AI and natural language processing. Today, its Watson supercomputer is continuing to innovate across a variety of industries, pushing the boundaries in areas like conversational AI, predictive analytics and natural language classification.
Microsoft has made several advancements in the larger world of artificial intelligence, from machine learning-enabled cybersecurity to cognitive computing. Most recently, the company has been collaborating with research firm Hugging Face, dipping its toes in the applied artificial general intelligence space. Hugging Face is known for its leading open-source library for building machine learning models, and plans to introduce its endpoints on Microsoft’s Azure. Together, the two companies plan to make significant inroads in AGI by fostering democratized machine learning strategies and open-source collaboration.
MindBridge’s Auditor tool uses AI to analyze and detect errors in financial data, determine levels of risks for transactions, and produce detailed risk assessments for financial institutions. While this may sound simple enough, Auditor is in fact making big leaps toward the ultimate realization of AGI. The machine intelligence-driven software was among the first to complete a financial audit with a totally automated workflow, according to the company, and its risk analysis system uses a variety of machine learning strategies that combine supervised and unsupervised learning. This is useful in not only the world of finance, but also medical research, corporate management and even national defense.
Backed by two decades’ worth of neuroscience research, Numenta is a key player in our understanding of how the human brain works, and has been at the forefront of several breakthroughs in the world of artificial intelligence. At the foundation of its technology is its Thousand Brains Theory of Intelligence framework, which helps the company to develop new architectures and algorithms that will be fundamental to advancing into artificial general intelligence.
Olbrain has a neural networks-based general intelligence platform that uses artificial theory of mind, a type of AI that can sense and respond to human emotions, to train robots. As far as Olbrain is concerned, the bots we are familiar with are old news — their intelligence degrades over time, their knowledge is rigid and they require a huge amount of data to run, making them inefficient. But Olbrain’s technology is working to create robots that have a generalized intelligence, where their efficiency does not decrease due to data drift, and that do need to be trained on huge amounts of data thanks to transfer learning, or the reuse of a pre-trained machine learning model on a new problem.
One Concern was created to help communities prepare for, respond to and recover from natural disasters, providing decision-makers with the data and analysis they need to make more informed decisions. The company has essentially combined AI, machine learning and human-centered hazard science to create a digital twin of the physical world, which reveals potential risks posed to our built and natural environments by extreme weather and climate change, whether that be to specific structures or external networks communities depend on to function. Advanced artificial intelligence and machine learning allow the platform to create intelligent, probabilistic models that are capable of learning, evolving and scaling from each new piece of data, according to the company.
OpenAI is a nonprofit research company on a mission to create artificial general intelligence, or AI that is capable of behaving and learning the same way humans can. Although AGI does not technically exist yet, OpenAI is one of the few companies to come close with the invention of GPT-3, an autoregressive language model that uses deep learning to produce human-like text. While it isn’t technically intelligent, GPT-3 has been used to create some pretty amazing things, including a question-based search engine and a chatbot that allows users to have conversations with historical figures. In the long term, OpenAI says it would like to continue building AGI safely, and has received backing from tech giants including Amazon, Microsoft and Elon Musk.