The Evolution of Artificial Intelligence
The tech we use nowadays is the outcome of a number of artificial intelligence milestones achieved by many unsung heroes. We travelled back in time and compiled a list of all significant artificial intelligence achievements that have allowed us to enjoy our current lifestyle. Let’s take a look at AI evolution over time. The timeline below highlights AI milestones that have been achieved throughout history. These achievements can be found in all areas of Artificial Intelligence. Because this is a growing timeline, new milestones will be updated on a regular basis.
Timeline
MIT researchers have created a self-sustainable microsystem
This research is funded by the U.S. Army Combat Capabilities Development Command Army Research Laboratory. This project is a perfect example of using a living system to generate electricity. Thus devised system won’t require any battery and can operate without any human input. The same research team discovered that electricity..Read more
Brain computer interface records brain signals for handwriting
Researchers at Howard Hughes Medical Institute set up a brain-computer interface to record brain signals when a person thinks of writing a letter. That was a unique idea to record and recreate the expression of a paralyzed person who wants to write. The neural signal for each letter was recorded..Read more
New Nex-Gen robotic Muscle Technology outperforms natural counterpart
Cavatappi artificial muscles work more like human muscles are built from polymer tubes that have the potential of twisting and coiling. These have the advantage over the actuators as actuators lack flexibility. They are very powerful and lightweight, outperforming human skeletal muscles. These twisted polymer actuators (TPAs) can contract up..Read more
Our adaptive immune system uses reinforcement learning against germs
The adaptive immune system is our body’s second line of defense. The soldiers are classified either as T cells or B cells. T-cells are responsible for detecting the foreign pathogen and creating a response against them. Previously, the biology of T cells has been widely studies but detection and training..Read more
Nanotech scientists create world’s smallest origami nanorobot
The field of robotics does not stop surprising us. Cornell University scientists used a memory actuator to create a nano-bot that folds itself with a volt supply. A shape-memory alloy is used i.e. bends on heat/current and regains its shape when heat/volt is removed. This concept has been utilized to..Read more
no-electronic robot born in University of California
Soft robotics is a new paradigm shift in robotic research. They offer flexibility and multi-environmental functionality, contrary to rigid robots. Most soft robots function based on pneumatic circuits (pressurized air and valves). But the main issue is their brain circuit that is usually heavier and requires electricity. Michael T. Tolley..Read more
Swinburne University Of Technology Australia Researchers Introduced Neuromorphic Processor
Researchers from the Swinburne University of Technology Australia and other continents collaborated to demonstrate the World’s faster optical neuromorphic processor with the capability of 10 TeraOPs/s. This is the fastest processing speed achieved by any single processor. While other ultra processors like Google’s TPUs can operate 100 Tera operations per..Read more
Milad Abolhasani developed Artificial Chemist 2.0
Remember Iron Man’s Jarvis? Artificial Chemist 2.0 works with the same charm. Milad Abolhasani from North Carolina State University demonstrated this phenomenon. Basically, it is a manufacturing system that enables users to select quantum dots for customized results. Quantum dots are colloidal semiconductor nanocrystals that are used in LED displays..Read more
Scientists Developed Soft Robotic Gripper Similar To Pole Beans
At the University of Georgia, scientists created a small and sensitive soft robotic gripper that gently but firmly holds the smaller objects, even a minimum of 1 millimeter in diameter. This idea was taken from pole beans that twine against a rope or any other elongated structure. The process of..Read more
DeepMind’s AlphaFold Won CASP Protein Folding Contest With Highest Accuracy
Protein structure prediction is the most crucial part of drug designing and understanding how life works. DeepMind’s AlphaFold was initiated to understand protein structure at its best. In a Critical Assessment of Structure Prediction (CASP) contest, AlphaFold competed against 100 competitors and scored 87 while 2nd top competitor was at..Read more
Waymo Self-Driving Cars Opened for the Public
It’s the very first autonomous car company to bring driverless cars to the public. These Waymo Self-Driving Cars are available in the greater Phoenix area, Arizona, with the full map. Users can download its app on their smartphones and call the taxi to the location. Previously, safety staff accompanied self-driving..Read more
Google Launches Its Fourth Generation Tensor Processing Unit (TPUv4)
Tensor Processing Unit is a unique Application-Specific-Integrated-Circuit (ASIC) specifically designed for AI acceleration in Neural Network Machine Learning. The software used in this structure is Google’s TensorFlow. Recently Google’s TPU has broken all previous performance records, measured in MLPerf benchmark competition. Google stood on top in 6/8 benchmarks, i.e., DLRM,..Read more
CurialAI, The First AI System To Detect Potential COVID-19 Patients Within One Hour
Oxford University designed an AI program, CurialAI, based on tests taken in hospital emergencies within one hour like blood tests and other vital signs. The core team members were Dr. Andrew Soltan from John Redcliffe Hospital, Professor David Clifton from “AI for Health” Lab, and Professor David Eyre of the..Read more
OpenAI released GPT-3 Beta
Generative Pre-trained Transformer-3 (GPT-3) is a powerful AI model for autoregression of natural language. It is considered the most amazing and most feared AI model in NLP, as it has the ability to generate human-like text models. Some antagonists believe that this model will pollute the textual quality over the..Read more
Facebook Incorporated GrokNet, A Computer Vision Model For Online Shopping
Groknet is basically a product recognition system used by Facebook Shops. This system utilizes advanced computer vision, image processing, and neural networks for determining the details of uploaded images. Groknet helps determine object categories like “pen,” Laptop,” basic attributes like object color, material, size, and any other related information that..Read more
Once-For-All Model Was Published By Han Cai And Team
Surprisingly, AI edge devices produce high carbon dioxide. A study revealed that training off-the-shelf language processing system emits 1400 pounds of CO2. Training full processing AI from scratch can cause 78000 pounds of emission. Han Cai and his team designed an efficient algorithm for training networks once for all, namely..Read more
AI Model Screened Out Halicin, A Strong Antibiotic Against Resistant Bacteria
Along with his team at MIT, Prof. James Collins used a machine learning model to identify a small uncovered compound from a drug used to treat diabetes. Artificial intelligence predicted that this compound could kill many bacteria with the least side effects on human cells. This AI model screened out..Read more
Microsoft launched its Turning Natural Language Generation (T-NLG)
Turning Natural Language Generation (T-NLG) is a natural language generator with 17 billion parameters (outperforming OpenAI’s GPT-2). It can perform tasks like freeform generation, answering the questions, autocompleting the sentences, and summarizing paragraphs. Read more February, 2020
Detectron2 Launched By Facebook
Facebook’s Detectron2 is an object detection platform that is implemented in PyTorch. Its previous version was implemented in Caffe2. It is now more extensible and flexible with high object detection algorithms like DensePose, panoptic feature pyramid networks, and the Mask R-CNN family. It can be trained on single or multiple..Read more
Textron Systems Launched Ripsaw M5 Autonomous Battle Tank
Ripsaw M5 Autonomous Battle Tank is an unmanned fully automatic electric war tank unveiled in the Association of the U.S. Army expo in Washington, D.C. It is the most advanced smart war wagon with multiple programmable capabilities, including customizable turret, 260-degree surveillance and situational awareness system, and routine clearing options..Read more
AI Outperformed Radiologist In Diagnosing Lung Cancer
Researchers from GoogleAI and Langone Medical Center, Center for Biological Imaging, prepared a deep learning algorithm that used lung cancer’s prior and currently computed tomography scans to predict lung cancer risks. In a comparison study, the AI outperformed radiologist algorithm competed with 6 radiologists to detect potential lung cancer through..Read more
FAISS By Facebook AI Research
Facebook AI Similarity Search (FAISS) is a library searching for similar multimedia content using efficient search algorithms. We know that multimedia files on social media exceed trillions of terabytes, and it is nearly impossible to find out similarities within the content using simple SQL engines. FAISS By Facebook involves a..Read more
GPT-2 Was Released With 1.5 Billion Parameters
Generative Pre-trained Transformer-2 (GPT-2) is a transformer machine learning model for auto-text generation. Using NLP and deep learning can perform various text-related tasks like answering questions, summarization, and translation. It has 1.5 billion parameters (training set). It works amazingly for short paragraphs but loses its senses for longer paragraph generation...Read more
Scientists At Politecnico Di Milano Designed A Feedback Circuit For One Step Algebraic Solutions
Linear Algebra is an integral part of almost all engineering sciences. Solving a matrix requires high computational power, while solving more complex calculations requires an extravagant setup. Matrix factorization is used in solving matrix equations, and it cannot be completed in a single step. Each step consumes energy and hardware..Read more
Waymo One Autonomous Car Launched By Waymo LLC
Waymo LLC is a self-driving car company under Alphabet LLC, a subsidiary of Google. In 2018, Waymo One autonomous car was launched in Greater Pheonix, Arizona. The core detection components were sensors, lidars, 360 surveillance radars that can detect an object 300 meters away. Waymo engineers developed a virtual driving..Read more
Facebook Started Using Computer Vision AI To Filter Out Explicit Visual Content
As the number of Facebook users grows, it is becoming hard to scan each post or data shared manually. Facebook started using computer vision AI and image processing to scan nudity and explicit graphics. Though there are many false positives but application is evolving day by day until it becomes..Read more
Lovot, First Emotional Partner Robot Created By Groove X
With more than 50 sensors, LOVOT can sense your mood changes and act accordingly to change your mood from upset to excited and peaceful. It has beautiful color-changing eyes and a warm body. It is more like a pet or a kid that is controlled through a mobile app. You..Read more
The First AI Robot CIMON Sent To Space
The First AI Robot CIMON is a head-shaped AI-mediated robot that was developed by Airbus and IBM. It is, in fact, Crew Interactive Mobile Companion. Fully voice-controlled, camera and repair attributes containing robot was destined to reduce the stress of astronauts. Read more June, 2018
OpenAI Built Generative Pre-Trained Transformer Model
Generative Pre-Trained Transformer is an NLP model for generative language modeling. Basically, its concept is based on the possibility that supervised learning can amalgamate with unsupervised pre-trained data sets for better and exponentiated results. GPT from OpenAI showed promising results and led the way for upcoming GPT-2 and GPT-3 models...Read more
Facebook AI Research Published StarSpace Algorithm
Facebook’s AI research division created StarSpace Algorithm for understanding user’s embeddings. User data on Facebook is highly unstructured, and analyzing such messy data is trouble for data scientists. By using user embeddings, this data can be classified and annotated efficiently. StartSpace helps classify text patterns, understand relational embeddings, and learn..Read more