Information systems
gmel
project.docxIt’s In The Details
· Generative AI uses a very sophisticated method of trial and error. Two networks collaborate in an effort to generate artificial candidates using a predetermined input and select the most genuine output. This process uses a platform known as Generative Adversarial Networks (GANs). The job of the first network is that of a generator. Depending on the task at hand this network can produce an image, text or other artificial medium using an authentic input. The second networks job is to filter the candidates, also referred to as discriminating, and select the most genuine outcome.1
Source: https://www.codemotion.com/magazine/dev-hub/machine-learning-dev/generative-ai-creating-objects-with-machine-learning/
· Over time both the generative network and discriminative networks learn from one another. Similar to a creator-critic relationship, the generative network learns the preferences of the discriminative network by means of back-propogration.2 This method of machine learning has allowed GANs to create new content that mirrors the original input to the point of plausible contention.
Significance
Many people may not realize it, but they are already using generative AI every day. Examples of daily use include sorting an inbox, editing photos, talking to chatbots, and using driving assist, to name a few. Generative AI is not a distant technology of the future, it is very much a technology of the present. However, the significance of generative AI in the future cannot be overstated and as businesses continue to tap into the massive potential of this powerful technology and apply it more readily, the clearer the impact will become.
Some of the current applications:
· When you call a bank or car dealership the first “person” you talk to is a chatbot. These are used to increase customer service and direct your call correctly.
· Looking through every email trying to find one would be extremely cumbersome. Generative AI helps sort the inbox according to parameters preset or set by the email’s owner.
· A growing area of generative AI is that of driver assist. Self-driving cars are all the rage but first came the driving assistance by keeping the car centered in the lane, alerting the driver of merging cars, and other safety measures.
· Some other working areas that use a great deal of generative AI are photo editing, facial editing, language translation, image understanding, targeted ads, security, audio development, and auto programming
So, what are the future impacts of generative AI? The answer is simple and widespread; anywhere where digital usage resides; generative AI can help to improve. These are some examples.
· A company wants to know what people think of their product and services, but it is too costly and time consuming to read through reviews. AI can scan all the reviews and create a summary of the current view and feelings towards products/services. This information is critical to businesses keeping the customers happy and staying ahead of the competition.
· As hackers become more developed, it becomes increasingly hard to protect important client and company information. Set generative AI into the software with the goal of defending the information and not only will it learn the patterns of the hackers but anticipate their future moves. Therefore, generative IT robots can learn the game of chess and beat grandmasters within a matter of hours.
· Models and actors could become a thing of the past. Generative AI is on the way to creating and developing human images by scanning hundreds of thousands of actual humans and then creating a digital version that is every bit as unique as each human is. Models and actors are expensive, but generating these digital beings is a cheap alternative that could be every bit as good as the real thing eventually.
· 3D printing has made leaps and bounds in recent years, but the widespread application of it has not quite arrived. Generative AI can help create better parts for cars, prosthetics, furniture, décor, and medical instruments. AMFG states: “Although still in its infancy, generative design has shown remarkable potential for industrial applications, particularly when coupled with 3D printing. Together, generative design and 3D printing can achieve more design flexibility, while creating lighter and stronger parts.”
As can be seen, generative AI has lots of current and potential applications. The examples used above are just a few of the numerous ways that the technology can be used to better lives, create productivity, increase sales, and cut costs. The next steps are for businesses to research and find ways to incorporate this powerful technology into their companies’ systems. As with many RDT&E projects, the upfront time and money spent will be more than worth it once the generative AI takes over and assists.
Projections
Generative AI has already become popular in the entertainment industry for creating previously presumed impossible content and enhancements. There are many uses for generative AI such as the following, that will continue to gain momentum and popularity because they are cheaper, faster, and as good if not better quality than organic creation methods: image process upscaling, augmenting VFX for movies and shows, restoring, preserving, and colorizing old films, and localizing and moderating original content.
There are many incredible possibilities and uses that generative AI can offer. One such use is the ability to use data from music of the past to generate new, original songs influenced by deceased musicians, such as Amy Winehouse, Jimi Hendrix, Curt Cobain-led Nirvana, and the Jim Morrison-led Doors5. Such technology will be able to restore not only existing media, but also create new media from past genius and its influence. However, entertainment uses are only elementary examples compared to the vast real possibilities of generative AI associated with synthetic data and populations.
Synthetic Data
What makes artificial intelligence so valuable is its speed and capacity for machine learning, also known as deep learning, and eventually, self-learning. “In order to train a deep learning model, researchers must collect thousands or millions of data points from the real world. They must then have labels attached to each data point before the model can learn from the data. This is at best an expensive and time-consuming process; at worst, the data one needs is simply impossible to get one’s hands on.”6 This data scarcity is due to numerous and increasing privacy restrictions and regulations on data, such as GDPR, CCPA and HIPAA.7 “Synthetic data upends this paradigm by enabling practitioners to artificially create high-fidelity datasets on demand, tailored to their precise needs. For instance, using synthetic data methods, autonomous vehicle companies can generate billions of different driving scenes for their vehicles to learn from without needing to actually encounter each of these scenes on real-world streets.”6
Synthetic Populations
Perhaps the most controversial element of synthetic data is the idea of creating and using synthetic populations of people to train AI systems. There are multiple elements to this idea. One element of synthetic data is creating digital twins of real people whose digital activities, behavior, decisions, and future decisions are not only known but silently influenced.8 These digital versions can be used in simulations for prediction purposes. Another element is generating a digital population of people who have never existed before, but act and make decisions exactly like our current population. The creation of this synthetic data is from a pool of real individuals rather than a specific real person, thus enabling AI to learn from datasets without compromising their privacy.7 “It is projected that by 2025, 10% of governments will avoid privacy and security concerns by using synthetic populations to train AI.”9 “As synthetic data approaches real-world data in accuracy, it will democratize AI, undercutting the competitive advantage of proprietary data assets.”6 Meanwhile, the core technology that can make synthetic data so beneficial and revolutionary, also threatens to have a potential massively caustic effect on society in the form of deep fakes.
Deep Fakes and Potential Drawbacks
Deep fakes have present-day beneficial value in the forms of localizing and moderating content for differing audiences. “Deep fake technology enables anyone with a computer and an Internet connection to create realistic-looking photos and videos of people saying and doing things that they did not actually say or do.”6 Such ability to create realistic yet, misleading depictions can create confusion and chaos in aspects of private privacy, domestic politics, diplomacy, and national security. There are methods to detecting deep fakes, but they are not as efficient as the capacity to generate content.10 The ability to decide what is authentic and what is fake, will soon be indistinguishable, cause major issues for a society based on the processing of facts. It is projected that “in 2023, 20% of successful account takeover attacks will use deep fakes as part of social engineering attacks.”9 The only solution to establishing communal trust may be stronger regulation of generative AI.
Factors in Projected Adoption
Further adoption of generative AI seems to be an inevitable, disruptive force that will change the landscape of how we create content and study patterns. However, there are barriers that will be tested, mostly in the form of public regulation and pushback. Content creators such as artists, writers, and actors may feel threatened by the quality and affordability of supplementing their work with that done by AI. Like with the invention of photography in 1839, art will adapt because of technology. Rather than creating replica art that was in competition with photography, other forms like cubism were developed. “Creativity transformed from copying to a higher level of expression. Here, AI is making creators experiment and making them push their own boundaries.”11
The largest hurdle critical to ensuring future adoption of generative AI is how to manage deep fakes. Regulations will need to be specific so that corrupt use of generative AI and the creation of unauthorized deep fakes can be eliminated or reduced without stifling the growth of beneficial aspects of data synthesis and deep learning. Cybersecurity consultants, Gartner, recommends “that organizations should schedule training about deep fakes. ‘We are now entering a zero-trust world. Nothing can be trusted unless it is certified as authenticated using cryptographic digital signatures.’”9 Such training is a start for private organizations, but the issue of deep fakes is already a public concern.
1. https://www.avenga.com/magazine/generative-ai/
2. https://athenasowl.tv/generative-ai-and-its-application-in-the-media-industry/
3. https://amfg.ai/2018/10/25/generative-design-3d-printing-the-manufacturing-of-tomorrow/
4.https://www.analyticsinsight.net/what-is-generative-ai-its-impacts-and-limitations/#:~:text=Generative%20AI%20refers%20to%20artificial,that%20to%20generate%20similar%20content.
8 - https://www.avenga.com/magazine/human-digital-twins/
9 - https://www.techrepublic.com/article/gartner-the-future-of-ai-is-not-as-rosy-as-some-might-think/
11 - https://athenasowl.tv/generative-ai-and-its-application-in-the-media-industry/
