Outlook 2023: Predicting future events in artificial intelligence

2022 has passed, and it is a special period of the year to look forward to future development. As usual, AI Era Frontier organizes a collection of articles about future technologies and trends for the benefit of readers.

In recent years, artificial intelligence has been growing at a rapid pace, and it seems difficult for anything to stop it. As momentum builds, which direction will AI take in 2023? Experts have something to say.

Many artificial intelligence projects are poorly conceived, leading to eventual failure. Zohar Bronfman, co-founder and CEO of Pecan AI, said that in 2023, enterprises will be more vigilant when evaluating the efficacy of artificial intelligence.

“In 2023, business leaders will evaluate potential data science projects more rigorously than in the past. Too often these projects fail to make a real impact because they are not aligned with business needs, or because they never make it to production. With the expense and time commitment involved in data science, leaders will review proposed efforts more carefully and pursue the right plan." Bronfman said: "We will continue to work to ensure that in the short term, based on the model's output, we can promote Business improvement actions, or stopping before resources are wasted."

In 2023, the demand for data scientists will continue to increase. Nick Elprin, CEO and co-founder of Domino Data Labs, predicts that demand for GPUs for deep learning model training will be the same.

“The biggest source of improvements in artificial intelligence has been the deployment of deep learning in training systems, especially Transformer models, tasks designed to simulate the actions of neurons in the human brain. These breakthroughs require enormous computing power to analyze large amounts of structured and unstructured data sets. Unlike CPUs, graphics processing units (GPUs) can support the parallel processing required for deep learning workloads. This means that in 2023, as more applications based on deep learning technology emerge, from From translating menus to treating diseases, demand for GPUs will continue to soar."

Supporting this view is Charlie Boyle, Nvidia’s vice president of DGX systems, who hopes to sell more GPUs next year.

"In 2023, inefficient, x86-based traditional computing architectures that cannot support parallel processing will be replaced by accelerated computing solutions that will provide the computing performance needed to build language models, recommendation engines, and more , scale and efficiency. Amid economic headwinds, enterprises will seek AI solutions that can achieve their goals while simplifying IT collaboration processes and improving efficiency. New platforms that use software to integrate workflows in infrastructure will enable breakthroughs in computing performance and reduce overall Cost of ownership, reduced carbon footprint, accelerated return on investment for transformative AI projects, replacing more wasteful, older architectures.”

How long do you think it will take to hire a qualified data scientist? Some people joke that it's as difficult as spotting a unicorn. Kyndi founder and CEO Ryan Welsh believes that 2023 will be the year when the world reaches "peak data scientist".

“The shortage of data scientists and machine learning engineers has been a bottleneck for enterprises to realize the value of artificial intelligence. As a result, two things have happened: more people are pursuing data science degrees and certifications, and the number of data scientists has increased. volume; and vendors coming up with new ways to minimize the involvement of data scientists in AI production. The simultaneous interference of these two trends leads to “peak data scientist.” Because as foundational models emerge, companies can build on those models Instead of requiring each company to train its own model from scratch, fewer data scientists will be needed to train custom models, while more of them are graduating. In 2023, the market is expected to React accordingly, leading to data science oversaturation.”

Triveni Gandhi, head of AI at data science tools provider Dataiku, predicts that we can expect to see ethical AI continue to attract corporate attention and resources.

“While we’ve seen some companies cutting ethical AI jobs in the news, the reality is that most companies will continue to invest in their ethical AI teams. This resource is critical to the scale and scope of AI Operations is critical, helping companies feel confident that their AI outputs are aligned with their values ​​and executed in a robust and reliable manner. Additionally, the Ethical AI group gives users confidence that the products they are interacting with have been considered and meet safety requirements and trust expectations. For any company to stay ahead of the curve, building an ethical AI team is a must.”

One of the dilemmas of deep learning is the black-box nature of predictive models. Jans Aasman, CEO of graph database maker Franz, said one way to solve this problem is to pair artificial intelligence with causal knowledge graphs in 2023.

“The next few years will see growth in causal AI, starting with the creation of knowledge graphs that discover causal relationships between events. Healthcare, pharmaceuticals, financial services, manufacturing, and supply chain organizations will combine domain-specific Knowledge graphs are linked to causal graphs and simulated to move beyond correlation-based machine learning that relies on historical data. Causal prediction has the potential to improve the explainability of AI by making causal relationships transparent."

Maya Natarajan, senior director of product marketing at graph database maker Neo4j, also foresees significant progress in combining graphs and AI.

Natarajan said: “Enterprises will continue to look for the best ways to leverage knowledge graphs to enable responsible artificial intelligence. By leveraging the context provided by knowledge graphs, organizations can improve the accuracy of ethical decision-making by keeping data stream sources to improve interpretability and help mitigate bias by opening up new analytical methods."

Next year artificial intelligence will find vector databases even more attractive. That’s what Edo Liberty, founder and CEO of Pinecone, one of the early leaders in the vector database market, thinks.

"As artificial intelligence continues to evolve and become more widely used, there will be a corresponding need for more advanced and scalable infrastructure to support its development and deployment. A key area for investment in artificial intelligence infrastructure Will be specialized data infrastructure, such as vector databases, designed to store and process the large amounts of data generated by modern ML models. Liberty said: "This will accelerate the development and deployment of artificial intelligence systems that will surpass the previous generation in many fields." One year of application. ”

In recent years, companies have been increasing their use of artificial intelligence, with mixed results. But Kimberly Nevala, business solutions manager at SAS Consulting, predicts that in 2023, artificial intelligence will enter a "less is more" growth Phase.

"As organizations realize that "less is more," AI will proliferate and quietly shift the focus from large-scale innovation as a goal to being applied to a wider range of small decision points and actions, Its collective impact is greater than the sum of its parts. Paradoxically, organizations and key employees need to have a broad understanding of these technologies and be comfortable using them. "

So you've invested heavily in the GPU to train your neural network. What do you do with it? There are always some SQL queries that require extra horsepower, said Matan Libis, vice president of product at SQream.

“The ability to reuse computing resources for AI/ML is an exciting and valuable opportunity for enterprises. Not only does reuse reduce the carbon footprint left by AI, but the general increase in cheaper global data storage solutions also reduces reliance on GPU hardware. Additionally, latency can be reduced when you don’t need to move data from one place to another. However, once enterprises prepare data in one place, train in another, and move inference to yet another, hopefully by streamlining the process, we will see huge improvements in the accuracy and speed of AI/ML capabilities. ”

Yonatan Geifman, CEO and co-founder of deep learning company Deci, said that the high cost of cloud computing is putting pressure on everyone, but artificial intelligence users can combat rising costs by optimizing models.

“Enterprises that have been running AI models in cloud environments are seeing the financial toll high-performance cloud processing can take on them. 2023 may see more companies looking to reduce AI inference cloud costs. One of the most effective ways to achieve this is to increase the speed of AI models while maintaining their accuracy, reducing processing time on the cloud and effectively saving money. ”

Evinced chief scientist Yossi Synett predicts that in 2023, we will see more breakthroughs in self-supervised machine learning technology that does not require labeled data.

“One factor hindering the development of artificial intelligence is the lack of high-quality labeled data. While we are already seeing progress today, growth will continue in 2023. We are finding more and more ways to use self-supervised learning to The model is pre-trained and then fine-tuned for the specific task. The best and most effective examples of this are NLP (Natural Language Processing), where it is called masked language modeling (making the model predict hidden words in a sentence) and causal The technique of language modeling (having a model predict the next word in a sentence) is a game changer. Since self-supervised learning does not require labeled data, fine-tuning requires much less labeled data, making it easier to train complex models. . can be used to better select labeled examples, which further reduces the financial barriers to AI projects."

Chintan Mehta, CEO and Group CIO of Wells Fargo, said to be prepared, Let artificial intelligence reach a higher level in 2023, adopt new user interaction models, and better understand intentions.

“In 2023 and beyond, the deployment of artificial intelligence and signal sensing will accelerate exponentially. AI will defeat biased perception, judgment and legal interpretation. The industry will build more solutions to break bias so that artificial intelligence Intelligence provides solutions to consumers while explaining their course of action. User interfaces will transform. They will move beyond app-based experiences from non-visual tap/touch interactions to contextually delivered visual action calls and language and gesture-based interactions. The artificial intelligence required to power these experiences will increase dramatically, going beyond just understanding language to truly grasp the hidden intent of each interaction. Artificial intelligence will beget artificial intelligence."

of German IT company GFT US CEO Marco Santos predicts that in 2023 we will see unprecedented use cases for artificial intelligence and machine learning emerge and eventually become mainstream.

“As companies break free from the constraints of legacy systems and are able to bring together massive data sets from disparate systems, we will see unprecedented use cases for artificial intelligence and machine learning. For example, in automotive manufacturing, We are just starting to see the emergence of next-generation manufacturing data platforms, or single and unified cloud-based platforms. Manufacturers are aggregating all the data across their entire organization. Once they have the data, they can start building AI applications."

The above is the detailed content of Outlook 2023: Predicting future events in artificial intelligence. For more information, please follow other related articles on the PHP Chinese website!