At the end of June, the global data analysis field was completely fried.
Two major data analysis companies, Databricks and Snowflake, have set their sights on large AI models. You must know that these two opponents usually play against each other often, and they often talk about performance, products and technology. But at this year's own conference, the two companies were surprisingly unanimous. They both defined the theme as: Data+AI, and announced that they will spend a lot of money to increase the AI model!
First, Databricks announced the acquisition of AI large-scale model startup MosaicML for $1.3 billion, the largest acquisition announced in the field of generative AI this year. Later, Snowflake announced its cooperation with NVIDIA to jointly focus on the development of AIGC and proprietary large models.
Some people may think that data analysis companies "participate" in the field of AI large models, either to add chaos or for capital market hype. However, in the author's opinion, behind data analysis companies investing heavily in large AI models, it is precisely after years of development and accumulation of big data and artificial intelligence technologies that digital intelligence productivity has gradually gained popularity. How to efficiently, safely and conveniently release data Intellectual productivity has become a must-answer question for thousands of industries in the era of digital intelligence.
The Essence of Data+AI: Unleash Digital Intelligence Productivity
What is the logic behind Databricks' huge investment in the acquisition of AI large models?
This is actually the general trend of the integration of Data+AI.
As we all know, in the era of digital economy, data has been recognized as a new core factor of production, while AI is regarded as a transformative production tool, and the combination of the two is expected to truly bring about a leap in productivity. The industry calls it the release of digital intelligence productivity.
However, it is no easy task to release digital intelligence productivity efficiently, safely and conveniently. With the growth of massive data, as well as the continuous iteration and evolution of data technology and artificial intelligence technology, users in various industries are also facing an extremely complicated situation:
First, data is becoming massive and diverse, and data analysis and various models are also becoming more complicated. Taking the OpenAI GPT large model as an example, the parameter scale of each generation of GPT models has increased exponentially in recent years. Today, hundreds of billions of large AI models are not uncommon in the market.
Now, training a large model not only requires huge hardware costs, but also requires a lot of energy in data processing, training and other links. This makes many users have the mentality of "wanting to use but dare not use" large AI models.
Second, the emergence of a large number of smart scenarios has reversely promoted the demand for more data training, reasoning, and analysis, which will increase the requirements for data processing, analysis, and other links. In the industry field, trying to access AI large models in various business scenarios will not only bring about the release of productivity, but also raise the requirements of data processing and other links to a new level. Data processing and analysis need to be automated and intelligent.
Third, the continuous expansion of data consumption groups has brought about an unprecedented situation of data consumption. In the past, data consumption groups were often a small number of management groups; now, data analysis and mining are required in a large number of business scenarios, which greatly promotes the increase of data consumption groups. For example, some joint-stock banks or large manufacturing companies in China already have more than 10,000 monthly active employees who consume data, and the proportion of employees continues to increase. In daily business scenarios, "using data" has been integrated into the work of various employees among.
In fact, in order to better help users release digital productivity efficiently, safely, and conveniently, data analysis companies have been accelerating the integration of Data+AI over the years. For example, integrating popular AI frameworks such as TensorFlow, supporting the development of machine learning tools, etc. Today, the layout of AI large models is more like a further evolution of the Data+AI trend, which is natural and logical.
So, what changes will generative AI or AI large models bring to data analysis?
First of all, the integration of AI will definitely make the way of data analysis more intelligent and convenient, and continue to lower the threshold of data consumption and use, while generative AI or AI large models will accelerate the intelligentization of data analysis, and will improve data analysis and intelligence. Revolutionize the way numbers are used.
For example, generative AI capabilities are integrated into many links such as query and retrieval, data cleaning and preparation, analysis and visualization, making data analysis extremely simple and convenient. Take the flow of data analysis requirements as an example. In the past, it was completed through the dialogue between people and the interaction process between people and the data platform GUI. Business personnel, data analysts and data engineers need to form a cycle from explaining requirements to feedback solutions , and is a preset process, the process is complex, inefficient and difficult to optimize and iterate.
After the integration of generative AI, it has really changed the past input and interaction methods, allowing data analysis to better fit the user's ideas.
Kyligence, a leading domestic big data analysis and indicator platform manufacturer, is an outstanding representative of the industry's earliest exploration of Data+AI. As early as 2019, Kyligence launched an AI enhancement engine, which can actively analyze business usage patterns based on actual data characteristics and query habits, so as to realize adaptive matching of data models to business query requirements. The work of data modeling, development and design becomes automated and intelligent.
Undoubtedly, the integration of generative AI and AI large models will bring better natural language understanding accuracy, thinking and reasoning ability, and natural language output. In addition to further accelerating data analysis to become intelligent, it will also completely affect data analysis. Analysis, data consumption, and data interaction will bring about transformative changes.
At present, both cloud service providers and data analysis companies recognize that generative AI is integrated into data analysis and is accelerating its deployment. It is reported that at the Kyligence User Conference to be held on July 14th, Kyligence will bring a blockbuster new product of Data+AI, aiming directly at the intelligent usage in the era of large models.
Secondly, the combination of generative AI or AI large models and data analysis platforms will make the industry's proprietary large model training and reasoning easier in the future, and the cost of large models is expected to drop significantly in the future.
At present, the training and reasoning of large models is still a complex and costly task. The training and R&D expenses cost millions of dollars, which makes many users complain. Lowering the threshold and cost will be a big deal for many users. .
Some experts believe that the combination of data analysis platform and large-scale model technology is expected to allow enterprises to use their own proprietary data to train and build generative AI models in a simple, fast and low-cost way, allowing users to have data control Easy development of custom AI models without rights and ownership.
It can be said that with the integration of generative AI, Data+AI is accelerating to open a new era, and the release of digital intelligence productivity is just around the corner.
How to Unleash Digital Intelligence Productivity: Look here, see you in Shanghai!
Well-known AI expert Wu Enda believes that with the popularization of open source AI algorithms, the key to the successful use of AI technology is data-centric AI (Data-centric AI). I take it for granted.
At the end of June this year, the $1.3 billion acquisition case fired the first shot for Data+AI. As generative AI and AI large models begin to be combined with data analysis, the core of Data+AI is still data-centric. Next, what new changes and new impacts will Data+AI bring to the interactive methods of data analysis, the use of enterprise data, and even the establishment of proprietary large models?
The 2023 Kyligence User Conference on July 14 deserves close attention. Founded by the founding team of Apache Kylin in 2016, Kyligence is currently a leading provider of big data analysis and indicator platforms, and has been recommended by Gartner for enhanced data analysis for three consecutive years.
Kyligence's layout in Data+AI shows that Chinese data analysis companies have always been at the forefront of industry changes, and they also have forward-looking explorations and layouts for the data analysis interaction revolution and intelligent use of data brought about by AI large models.
It is reported that at the 2023 Kyligence User Conference, Kyligence will release a blockbuster new product of Data + AI. In addition, this conference will also gather industry leaders, technical experts, business executives and partners at home and abroad. In addition, experts from CICC, Debon, Ping An Bank, China CITIC Bank, Sany Heavy Industry, Amazon Cloud Technology and other fields will share many important contents in the field of Data + AI.