China
Africa
Explore chapters and articles related to this topic
Artificial Intelligence and Machine Learning
Published in Yves Caseau, The Lean Approach to Digital Transformation, 2022
The case of DeepMind is particularly interesting because the vision of its founder, Demis Hassabis, is precisely to build complex cognitive systems by assembling multiple methods.10 This approach has given AlphaGo and AlphaZero (capable of learning Go and chess by playing only against oneself) and also AlphaFold, a software that applies deep learning to protein structure reconstruction, one of the most important scientific problems of the time, with major impacts to help design new drugs. AlphaFold is today the most powerful algorithm in this field11 and another illustration of the use of deep learning as a component in a hybrid approach. In the book Human + AI, the authors give examples of enriching intelligent assistants with other AI techniques, from NLP tools to deep learning. The development of Google Duplex comes to mind here, but the authors detail the example of SparkCognition: “a product called DeepArmor, which uses a combination of AI techniques including neural networks, heuristics, data science algorithms, and natural language processing methods to be able to detect previously unseen threats.” Another very interesting form of hybridization is the combination of random exploration to generate data with AI. These “generative” approaches allow new areas of design to be explored. The machine is able to produce, select and then optimize new objects, whether they are designs—an approach that has been strongly emphasized by Autodesk—or processes.
Bottom Up Speech Recognition
Published in Robert H. Chen, Chelsea Chen, Artificial Intelligence, 2022
In a first step, AlphaFold employs a deep neural network to extract features from a training dataset and then searches for plausible protein structures having those features. It compares a protein's amino acid sequence with similar ones in the training set to find pairs of amino acids that appear in tandem, but do not lie next to each other in a chain, implying that they are positioned near each other in a folded protein in a process it called Multiple Sequence Alignments. The DNN was trained to take the pairings and predict the distance between them in the folded protein. Then the predictions were compared to precisely measured distances in known proteins and thereby enabled realistic guesses on how the proteins may fold.
Gathering the Team
Published in Volker Knecht, AI for Physics, 2023
In 2020, AlphaFold by DeepMind made a breakthrough in protein folding, a “holy grail” problem in the field of biology. Here the task is to predict protein structure from the sequence of amino acids. AlphaFold achieved an accuracy regarded as comparable with experimental techniques.4,5 Providing this software, AI thus solved one of the hardest problems in science.
Technology fitness landscape for design innovation: a deep neural embedding approach based on patent data
Published in Journal of Engineering Design, 2022
Prior studies have pointed out that technological improvement or novelty arises from the recombination or synthesis of existing technologies (He and Luo 2017; He et al. 2019), which, in our cases, can be viewed as such mutations of the technological genotype. Following the analogy framework, the latest innovations in autonomous vehicles have changed the genotype of automobiles and increased the values of automobiles by fusing artificial intelligence to assist or automate driving and battery-powered electric powertrain to replace combustion engines. Similarly, recent progress on the structure prediction component of the ‘protein folding problem’ achieved by DeepMind also presents the power of incorporating deep learning techniques (AlphaFold) into traditional biological domains (Jumper, Evans, and Pritzel 2021). In the past, it would take biologists six months to predict a protein structure, while now it takes only a couple of minutes using AI. Speaking in biological evolution terms, these domains’ genotypes have been mutated with increased fitness in the total technological space. The new genotype is positioned closer to the global peak in the technology fitness landscape.
Water resource prospects for the next 50 years on the water planet: personal perspectives on a shared history from Earth Day, the Fourth Industrial Revolution and One Health to the futures of alternative energy, bioconvergence and quantum computing
Published in Water International, 2021
Zooming in at the other end of the spectrum and with a view towards changes over time going forward, Google-owned DeepMind Technologies has developed an AI program known as AlphaFold which has recently outstripped numerous others in an annual competition to accurately predict the three dimensional shapes of proteins (Callaway, 2020). This convergence of technologies to make such tremendous strides in the prediction of three-dimensional structures with broad biological implications may be the closest AI has come to human intelligence thus far. Related advances, such as in the evaluation of aquatic ecosystem health through ‘culture-independent interconnected meta-omic approaches’ and other ‘high-throughput molecular technologies’, are anticipated to have far-reaching applications in wastewater treatment, bioremediation and numerous other areas of endeavour essential to sustainable development (Michán et al., 2021, p. 870).
Battling COVID-19 using machine learning: A review
Published in Cogent Engineering, 2021
Krishnaraj Chadaga, Srikanth Prabhu, Bhat K Vivekananda, S. Niranjana, Shashikiran Umakanth
drugs that can work for different protein structures. However, finding this unique 3D structure takes a lot of time. Evaluating genetic sequences and protein structure can be simplified with the help of ML and DL systems. AlphaFold was introduced by Google DeepMind (HospiMedica, 2020b; Institute for Protein Design, 2020), an advanced ML application that predicts the function of 3D proteins using their genomic sequences. The system can be used for COVID-19 as well. It also helped the scientific community to understand the virus by publishing a protein prediction for all the proteins that are related to SARS-CoV-2. The first 3D atomic map to understand the spike protein component of the virus was developed by the researchers at the University of Texas. AlphaFold has provided accurate predictions for the protein spike structure. 3D atom models for the SARS-CoV-2 protein scale that were simulated (Rees, 2020) was used by the researchers of the University of Washington. New proteins were created to emulate Coronavirus. These proteins would embed with the healthy cells to prevent viral attacks. AI was combined with cloud computing to prevent the spike protein from binding to healthy human cells. A vaccine could be produced using this technique (TABIP, 2020). The researchers of Flinders University studied the structure of Coronavirus and then designed a vaccine using the data collected (Flinders University, 2020). It was called Covax-19 as shown in Figure 4.