Original | Wen BFT robot
01
The development of artificial intelligence ushered in a new inflection point
ChatGPT is awakening global awareness of the transformative potential of artificial intelligence (AI), unleashing an unprecedented wave of attention and creativity.
The technology can mimic human conversation and decision-making, bringing us to the first real inflection point in public adoption of artificial intelligence. Ultimately, the disruptive potential of this technology will be tangibly felt by everyone, everywhere. Just two months after its launch, ChatGPT reached 100 million monthly active users, making it the fastest growing consumer app ever.
Base model is a general term for large models that have billions of parameters. Recent advances have enabled companies to now build specialized image and language generation models from these base models. The large language model (LLM) is both a generative artificial intelligence and a basic model.
The large language model behind ChatGPT marks an important turning point and milestone in the development of artificial intelligence.
Big language models are changing the rules of the market with two advantages. First, such models crack the code of language complexity. Today, machines have the unprecedented ability to learn language, contextual meaning and expressive intent, and independently generate and create content. Second, after pre-training with large amounts of data (text, images, or audio), these models can be adjusted or fine-tuned for many different tasks. This allows the user to reuse the model as it is or with minor modifications in a variety of ways.
Business leaders have generally recognized the importance of this moment.
They foresee how big language models and generative AI will fundamentally change business, academia, and society itself, opening up new frontiers of capability. These new technologies have had a huge positive impact on human creativity and productivity. Accenture research, for example, found that 40% of work time across all industries will be assisted by large language models such as GPT-4.
This is because language tasks account for 62% of the total working time of enterprise personnel, and 65% of the time can be improved with the help of human enhancement and automation technology to improve the productivity of work activities.
02
Development history: Milestones in the development of generative artificial intelligence
Machine Learning: Analysis and Prediction Phase
In the first decade of the 21st century, various types of machine learning technologies developed rapidly to analyze massive amounts of online data, draw conclusions from the output information, or learn.” Since then, enterprises have regarded machine learning as extremely important. The powerful field of artificial intelligence for analyzing data, discovering patterns, generating insights, building predictions, and automating tasks at a speed and scale far beyond our reach.
Deep Learning: Stages of Visual and Speech Processing
Entering its second decade, artificial intelligence has made great strides in its perception capabilities, a field of machine learning known as deep learning. During this time, breakthroughs in deep learning have been made. On the one hand, the realization of computer vision is helpful for the classification and detection of objects by search engines and self-driving vehicles; at the same time, it can also support speech recognition, so that widely used artificial intelligence voice assistants can interact with users in a more natural way .
Generative AI: Entering a new phase of language mastery
Based on the exponential growth in scale and capabilities of deep learning models, the next decade will be the era of machines mastering language. The GPT-4 language model developed by OpenAl marks a new functional stage for language-based artificial intelligence applications. Models such as these will have a profound impact on business because language is inextricably linked to all aspects of the day-to-day work of a business - institutional knowledge, interactions and operational processes all depend on it.
03
Use or Customize: Popularization and Applications of Generative AI
A series of easy-to-use generative artificial intelligence applications such as ChatGPT, Wenxin Yiyan, Tongyi Qianwen 34DALL·E, StableDiffusion, etc., are rapidly promoting the popularization of technology in the business field and the public, which will have a profound impact on enterprises Impact.
Because the large language model has the ability to process large-scale data sets, it can "grasp" all the information accumulated by the enterprise for a long time, including the development history, development background, business characteristics and business intentions since its establishment, and even down to products, markets and customers. . All content conveyed by language records, such as applications, systems, documents, emails, chats, video and audio, etc. will be innovated, optimized and reshaped, and finally reach a new height.
97% of the global executives surveyed believe that the basic model of artificial intelligence will realize the interconnection of cross-data types, which will completely change the use of artificial intelligence and the way it is used.
We are entering the next stage of the technology adoption cycle, and most enterprises will start to implement business applications by purchasing "models as a service". For many businesses, though, the greatest value comes from using their own data to customize or fine-tune the model to meet their unique needs.
use
Generative artificial intelligence and large language model applications are now readily available and usable at any time. Businesses can invoke these programs through application programming interfaces (APIs) and use prompt engineering techniques such as prompt tuning and prefix learning to customize them to a lesser extent for their specific needs.
custom made
But most businesses need custom models, fine-tuned with their own data, to expand their use and value. This enables the model to support some specific downstream tasks across the business. By doing this, enterprises can effectively use artificial intelligence to achieve a leap in performance—improving employee capabilities, improving customer satisfaction, introducing new business models, and timely sensing upcoming changes.
04
Looking ahead to ever-changing technology, regulation and business
Businesses will use these models to reshape the way work is done. As employees working alongside AI sidekicks become the norm, every role in every business has the potential to be completely reinvented, dramatically expanding what humans alone can achieve. In any given job, some tasks will be automated, some will be assisted, and some will be largely technology-agnostic.
Beyond that, a host of new tasks await humans, such as ensuring the accurate and responsible use of new artificial intelligence systems.
Enterprises should pay special attention to the impact of artificial intelligence on the following positions:
Content Creation:
Generative AI will become an essential creative partner, not only revealing new ways to reach and engage audiences, but also in fields such as production design, design research, visual identity, name proposal copy generation and testing, and real-time personalization. Unprecedented speed and innovation.
Enterprises are introducing the most complete artificial intelligence system DALL·E for social media promotion. DALL·E creates realistic images and artworks based on text descriptions, and can process up to 12 billion parameters when converting text to images, which can be shared on Instagram and Twitter.
Write code:
Software coders will use generative artificial intelligence to greatly improve productivity - quickly convert one programming language to another, master various programming tools and methods, realize coding automation, predict and prevent problems, and manage System documentation.
Accenture is trying to use the OpenAl large language model to improve developer productivity by automatically generating documentation. For example, clarify the rationale for SAP system configuration and set various functional or technical parameters. This solution enables users to submit requests through Microsoft Teams chat conversations while they are working; a properly assembled document is then quickly returned—a prime example of how specific tasks can be enhanced and automated, without changing the entire job.
automation:
Generative AI's sophisticated understanding of historical context, next best action, summarization capabilities, and predictive intelligence will enable a new era of hyper-efficiency, hyper-personalization, and automation of business processes in both back-office and front-office environments Take it to a new level of transformation.
A global bank is using generative artificial intelligence and large language models to transform the way it manages its volume of post-transaction emails, such as automatically drafting messages with action recommendations and sending them to recipients. This not only reduces the workload, but also allows for smoother communication with customers.
Security:
Over time, generative AI will support enterprises to strengthen governance and information security, prevent fraud, improve regulatory compliance, and proactively identify risks by establishing cross-domain connections and inference capabilities both inside and outside the organization.
In strategic cyber defense, large language models can serve a variety of useful functions, such as interpreting malware and quickly classifying websites. ? But in the short term, businesses are likely to see hackers exploit the strengths of generative artificial intelligence to generate malicious code or craft the perfect phishing email.
Moments like these are not common at present. Investments in generative AI, large language models, and foundational models will be enormous over the next few years. Unlike in the past, technology, regulation, and business applications will evolve in parallel, and at an increasing pace. In previous innovation curves, technological development has generally outpaced adoption and regulation.
technology stack
The complex technologies underpinning generative AI are expected to evolve rapidly at every level of the stack, with wide-ranging business implications. The amount of computing required to train top AI models is growing exponentially—it is now doubling every 3.4 to 10 months, according to various reports. As a result, cost and carbon footprint have become core considerations for adopting energy-intensive generative AI.
"The hottest new programming platform is the napkin." Paul Daugherty -- Global President and Chief Technology Officer, Accenture Technology Services
He was referring to the fact that entrepreneurs are using OpenAl to create job sites based on ideas sketched on napkins.
Risk and Regulatory Environment
Businesses will have thousands of ways to apply generative AI and underlying models to maximize efficiency and enhance competitive advantage. It's clear that companies are gearing up for this new track. Enterprises need to start from an overall strategy. In addition to generative artificial intelligence and large language models, all types of artificial intelligence and related technologies that they intend to use must be fully considered.
ChatGPT has further triggered people's thinking on the healthy development and normative application of artificial intelligence. Businesses in particular need to pay close attention to any legal, ethical and reputational risks they may face when technology is evolving and adopted faster than legislation is being made.
It is important to note that generative AI technologies, including ChatGPT, are designed with accountability and compliance in mind to ensure that such models and applications do not pose unacceptable risks to the business.
As an industry leader in responsible artificial intelligence, Accenture defined and implemented the principles of responsible artificial intelligence as early as 2017, and then integrated them into our business practices and client services. Responsible AI is the practice of designing, building, and deploying AI following clear principles to empower business while maintaining the public interest and benefiting society. Enterprises can also give full trust to artificial intelligence and expand the use of artificial intelligence with confidence.
AI systems need to be “perfected” with diverse and inclusive input data sets that embody broader business and social responsibility, fairness and transparency. If artificial intelligence can be designed and implemented within an ethical framework, it can accelerate the realization of the potential of responsible collaborative intelligent tools that closely integrate human intelligence and intelligent technology.
This not only creates a foundation of trust for consumers, professionals and society at large, but also improves business performance and opens up new sources of growth.
The Scale of Enterprise Adoption of Generative AI
To create value from AI, companies must transform the way they work across the board. Business leaders need to start now, designing jobs and tasks in new ways, and reskilling people. Ultimately, every role in the enterprise is likely to be reshaped, with today's work broken down into a set of tasks that can be automated or augmented by artificial intelligence, and the future of human-machine collaboration reimagined.
As we learn more about generative AI, it will disrupt traditional ways of working and usher in a new era of human-AI collaboration. Most workers will have powerful “assistants” that will fundamentally change how and what work is done.
Almost all jobs will be affected, and many new jobs will continue to emerge. Companies that take immediate steps to break down work into tasks and invest in training people to work with machines in a different way than they have done in the past will be able to achieve leaps and bounds in performance far beyond short-sighted rivals.
Nearly six out of ten businesses intend to use ChatGPT for learning purposes, and more than half plan to conduct pilots in 2023. More than 40% of enterprises are willing to invest heavily in this.
05
Actively welcome the era of generative artificial intelligence: six key points of technology application
business driver
Even with the benefits of innovative technology, rolling it out across an organization can be difficult, especially when the new technology revolutionizes existing ways of working.
Enterprises can first try the many functions of generative artificial intelligence, accumulate early success and get the support of change advocates and opinion leaders, continuously increase the acceptance of new technologies among employees, create the necessary conditions for further adoption, and then initiate transformation and skills Retraining agenda.
Businesses must try both. One, focus on low-hanging fruit opportunities, using consumable models and applications to realize returns quickly. Second, focus on using models tailored to your own data to reshape business, customer engagement, and products and services. Business-driven thinking is key to defining and successfully establishing application patterns.
As enterprises carry out various explorations of artificial intelligence to reshape their business, they will actually gain value, clarify the most suitable type of artificial intelligence in various application scenarios, and clarify the scale and complexity of investment. They can also test and improve methods around data privacy security, enhancing model accuracy, preventing bias, protecting fairness, and knowing when safeguards “human in the loop” are needed.
98% of global executives agree that AI-based models will play an important role in their corporate strategy in the next three to five years.
A bank uses an enhanced search tool to provide employees with the information they need
A large European banking group uses the Microsoft Azure cloud platform and the GPT-3 large language model to help employees with electronic document retrieval. This move allows users to get answers to their questions quickly, saving significant time and improving accuracy and compliance.
To further improve staff skills, the bank has built a three-year innovation plan, which will later apply generative artificial intelligence in areas such as contract management, conversational reporting, and bill classification. This move not only upgraded the internal knowledge base and helped employees obtain the information they needed, but also helped advance their goal of becoming a data-driven institution.
people oriented
For generative AI to succeed, companies need to focus as much on people and training as they do on technology. Therefore, they should substantially increase their investment in talent to address two distinct types of challenges: creating AI and using AI. This means developing talent in technical capabilities such as AI design and enterprise architecture while training people across the organization to work effectively with AI-enabled processes.
For example, in our analysis of 22 job categories, we found that large language models affect all categories from a minimum of 9% to a maximum of 63% of each workday. In 5 of 22 occupations, large language models could leverage large language models to revolutionize more than half of all work hours.
In fact, an independent economic study shows that companies are woefully underinvesting in helping workers keep pace with AI developments, which require more integrated cognition and judgment-based task-setting. Even domain experts who are proficient in how to apply data in the real world (e.g., doctors interpreting patient health data) lack sufficient technical knowledge to understand how these models work and trust technology as a "working partner"
Companies will also create entirely new roles, including linguistics specialists, AI quality controllers, AI editors and prompt engineers. For the most promising areas of generative AI, companies should start by breaking down existing jobs into basic task combinations. Then assess the extent to which generative AI might impact each task—complete automation, human augmentation, or nothing to do with it.
Prepare proprietary data
In order to customize the base model, enterprises need to use domain-specific enterprise data, semantics, knowledge and methods. Before the era of generative AI, enterprises could derive value from AI through an application model-centric approach to AI without modernizing their data architecture and assets. However, things are very different now. The underlying models require large amounts of carefully organized data to learn, so cracking the data challenge has become an imperative for every business.
Organizations need a strategic, disciplined approach to acquiring, developing, refining, securing and deploying data. Specifically, a modern enterprise data platform should be built on a cloud environment, which includes a set of reliable and reusable data products. With the cross-functional nature of such platforms, enterprise-grade analytics tools, and storage of data in cloud warehouses or data lakes, data can be untethered from organizational silos and used pervasively across the enterprise. Enterprises can then centrally analyze all business data at one location or through a distributed computing strategy such as a data grid.
Invest in building a sustainable technology base
To adequately meet the large-scale computing demands of big language models and generative AI, companies need to consider whether they have the right technical infrastructure, architecture, operating model, and governance structure, while paying close attention to costs and sustainable energy consumption. They must try to evaluate these technologies against other AI or analytics tools that might be better suited for a particular application model and cost a fraction of the cost, both from a cost and benefit perspective.
As the use of artificial intelligence increases, so will the carbon emissions from the underlying infrastructure. Therefore, companies need to establish a strong green software development framework that considers energy efficiency and material-related emissions at all stages of the software development life cycle. AI can also play a broader role in making business more sustainable and meeting environmental, social and governance (ESG) goals. Our research found that 70% of companies that have successfully reduced emissions in production and operations have used artificial intelligence.
Accelerate Ecosystem Innovation
Creating the base model is likely to be a complex, costly and computationally intensive endeavor. Almost all but the world's top organizations cannot accomplish this task on their own, it is beyond their capabilities and methods. It’s exciting to note that businesses can now tap into the power of an emerging ecosystem, thanks to massive investment from hyperscale cloud providers, tech giants, and start-ups. Global investment in AI startups and growth-stage companies is expected to exceed $50 billion in 2023 alone. “These partners can bring best practices honed over the years and provide valuable insight into how to efficiently and effectively use the underlying model in a specific application model. Having the right network of partners—including technology companies, professional service providers, and academic institutions , will be the key to navigating rapid change.
Improve your own level of responsible artificial intelligence
The rapid adoption of generative AI presents a new urgency for all enterprises: establishing a robust and accountable AI compliance system. This involves two things - establishing control processes to assess potential risks in the way generative AI is applied during the design phase; and developing clear steps to embed a responsible AI approach across the business. Accenture research shows that most businesses still have a long way to go. Our 2022 survey of 850 global executives shows that respondents generally recognize the importance of responsible AI and AI governance. But only 6 percent of companies believe they have built a sufficiently robust foundation for responsible AI.
Enterprise principles for responsible AI should be defined and led at the top and translated into an effective risk management and compliance governance structure, including organizational principles and policies, as well as applicable laws and regulations. The responsible use of AI must be led by the CEO, starting with enhanced training and awareness, and then expanding to focus on enforcement and compliance. Accenture pioneered this approach to governing responsible AI years ago, setting a CEO-led agenda and now taking it a step further with a formal compliance program. Our own experience shows that a principles-driven approach to compliance provides guardrails, yet is flexible enough to update as technology rapidly evolves, ensuring that businesses are not constantly struggling to “catch up”.
To be accountable by design, companies need to move from a reactive compliance strategy to proactively developing sound and responsible AI systems. And this must be done through a comprehensive framework that covers: principles and governance measures, risk management, policy and controls, as well as technology, enablers, culture and training.
Timing is everything. In a recent technology trend survey by Accenture, 72% of the 225 interviewed Chinese corporate executives were very, or extremely excited about the new functions brought by the basic model of artificial intelligence, but the proportion was slightly lower than the global average, and there is room for Further exploration of the potential and applications of generative artificial intelligence. For the benefits of artificial intelligence big language models, Chinese companies are more positive than global expectations in areas such as rapid and large-scale analysis capabilities, improvement of employee skills, development of new artificial intelligence applications and services, communication, processes and talents, but are accelerating innovation. , improved customer experience and quick decision-making expectations are lower than global.
Nevertheless, more and more Chinese enterprises are actively exploring generative artificial intelligence technology and starting to apply large-scale language models to achieve more innovation and efficiency improvement. For this purpose, we have sorted out the methods and application suggestions suitable for local deployment of Chinese enterprises.
In China, there are three main ways to apply large language models: SaaS, private cloud deployment, and localized deployment.
At present, the SaaS-based deployment method is the most mature, with the foreign Azure OpenAl service as the benchmark. But in the domestic market, Baidu's Wenxin Yiyan and Ali's Tongyi Qianwen are participating in fierce competition. Compared with the services provided by Azure, the SaaS services provided by domestic manufacturers have more advantages in data security and compliance, although the comprehensive capabilities still need to be strengthened.
Serving relatively professional customers, making full use of the industry knowledge provided by customers while ensuring that it is not available to competitors. The more flexible server usage strategy also greatly reduces the upfront investment of this method compared with localization. On the whole, this is the most feasible implementation method for domestic vertical industry customers.
There are many options for localized deployment. The academic community provides ChatGLM of Tsinghua University, Alpaca provided by Stanford, and commercial companies provide Dolly of Databricks, large language model of image specialization of Scale.ai, etc. Compared with the above two methods, the localized deployment method has two problems, such as high cost and uncertain use effect. Therefore, it is currently in a very early stage, and whether it can be used further remains to be seen.
In general, large language models are in a stage of rapid development, and their future form cannot be predicted, but what is certain is that large-scale applications must be an inevitable trend. No matter in the fields of scientific research, business or civilian use, large language models have broad application prospects, and the continuous innovation and progress of technology also provide a broader development space for its future applications.
Enterprises need to continue to invest in continuous development of business operations and personnel skills training just like technology investment. Completely reimagining the way work is done and helping employees keep up with technology-driven changes will be two of the most important factors in realizing the full potential of AI technology leapfrogging.
At present, Chinese enterprises are in a critical period of making breakthroughs in artificial intelligence. Artificial intelligence can not only reshape enterprise business, but also change entire industries. The future is promising.
Report source: Accenture
Report Editor: Intelligent Robot System
For more exciting content, please pay attention to the official account: BFT Robot
This article is an original article, and the copyright belongs to BFT Robot. If you need to reprint, please contact us. If you have any questions about the content of this article, please contact us and we will respond promptly.