Artificial Intelligence has gradually transcended the realm of science fiction to become an everyday tool. It has quietly shored up its presence in every corner of our lives, from streaming recommendations to supply chain optimization. At the same time it has become a recurring topic of conversation, gaining ever greater breadth and depth, from its passionate admirers to its most fervent detractors.
This highlights the importance of taking a closer look at the multiple applications of AI in Industry and better understanding the key considerations for conducting responsible regulation to enable optimal implementation of this technology in the context of digital transformation and glimpses into the beginning of a new industrial revolution.
Let’s explore together its real scope and the perspectives that will mark its path in the near future.
An Everyday Companion
Although generative AIs, such as Bard or Vertex AI, put Artificial Intelligence in the spotlight, the truth is that we have been living with this technology for a long time now, but often when the technology is successful it tends to become invisible. That is why it is important to highlight those AIs or their functionalities with which we coexist almost without realizing it.
One of the clearest examples are intelligent assistants, such as Google Assistant, on our smart devices, mainly mobile, an AI capable of executing different functionalities such as setting alarms, initiating searches, saving notes and more.
You may be interested in: The evolution in data engineering
However, this is just one of a long list of examples that we use in our daily lives, another more discreet, but effective artificial intelligence is the one behind the Google search engine, which is able to understand our search intentions, offering the best answers and anticipating possible questions in addition to displaying related searches.
Something similar happens with the search engines and recommendations of music and video streaming platforms, being able to create playlists based on our musical preferences according to our latest reproductions.
This thanks to hyper-realistic results and unethical implementation. Originating controversy over how AI should be regulated, even the joint strike between the Hollywood writers and actors guild responds to concerns over relevant regulations to ensure that the adoption of AI in the industry does not represent negative repercussions.
Explore these resources: Get to know the 7 V’s of Big Data.
This is a very delicate issue, since factors such as copyrights, the responsible use of image rights, and the adequacy of jobs to avoid massive staff cuts must be taken into account.
This concern is not only limited to the use of AI to replace or substitute actors or writers, since with the implementation of generative AIs, millions of people are strongly expressing their concern about the future of their jobs. Another example of these concerns is the discussion of the use of AI in art, since it implies rethinking the concept of the artist, his technique and the scope of AI.
Similarly, there are corresponding concerns in other sectors and industries where the reliability of this tool is questioned, such as in the security sector, where the so-called “hallucinations” have called into question the effectiveness of AI, given the failures produced by it.
Hallucinations, and other concerns, what happens when AI fails
The detractors of AIs have not missed a single failure of this tool, but why do they occur, and what happens when they fail, to answer these questions it is pertinent to understand a little more about how AIs operate.
While we can attribute some failures in their operation to their improper use, another large part of the errors is directly related to the training behind the AI, giving rise to the so-called “hallucinations”.
In general terms, a “hallucination” is an erratic response by the AI itself that can range from creating an erroneous response by not understanding the request entered, to creating or distorting information with data that appears to be valuable, but in reality turns out to be inaccurate or outright false. In an effort by the AI to try to provide an effective response to a request that is misunderstood or exceeds its limitations.
You might be interested in: Technology revolutionizing legal spaces, Legatech IA
Hallucinations in AI can arise due to a variety of reasons, such as errors in the programming code, the presence of incomplete and/or contradictory data, as well as the difficulty of the AI to fully understand the context. In addition, constraints in data availability and in the training process are also contributing factors to the occurrence of “hallucinations”.
To understand it better, it is essential to have a synthetic vision of the path of Machine Learning, a process underlying the creation of an AI of which we can highlight the following points:
- Problem Definition: What Challenge Do We Want to Solve?
Let’s imagine that an e-commerce company wants to predict the demand for its products in order to optimize its inventory. Here, the problem arises: How can we use AI to predict how many products will be sold in the next month? - Data Collection and Preparation: What Data is Relevant?
The company collects historical sales data, information on promotions, special events, and seasonality. This data is organized and prepared for use in the ML model. - Model Training: How Does AI Learn?
Using the data collected, the AI learns patterns and relationships. For example, it may discover that sales increase during the vacation season or during special discounts. - Evaluation and Adjustment: How Accurate is the Model?
The model is tested with new data to assess its accuracy. If the sales prediction does not match reality, the model is adjusted to improve its performance. - Implementation: How Do We Make AI-Based Decisions?
With the model trained and tuned, the company begins to use it to make sales forecasts. This helps them plan their inventory more effectively. - Continuous Validation: How Do We Avoid AI Hallucinations?
As the model is used, the company should be alert to possible hallucinations. For example, if the model predicts unusually high demand, it should be checked for unexpected factors causing it.
When there is no adequate and continuous monitoring during the training process, hallucinations tend to become more frequent and response accuracy will decrease. This is why the training and data curation process is so important.
On a technical level we already have a better idea of what happens when AI fails, but what are the implications of this on a day-to-day basis and how can we deal with it?
This is a question that implies new levels of complexity, since some failures may be inconsequential, while others may have serious repercussions; therefore, identifying the framework in which an AI error occurs is paramount. This also highlights the importance of establishing the primary legal basis for regulating the appropriate use of AI as well as its scope and limitations in different sectors.
To make this point I would like to revisit the incident involving Porcha Woodruff, a nursing student in Detroit, who was arrested and handcuffed at her home during a police raid on charges of robbery and carjacking. This after authorities used an AI facial recognition tool on a video of the incident.
However, as you may already know or assume at this point, that identification turned out to be erroneous. Woodruff, however, faced the possibility of going to prison while she was eight months pregnant, a factor that helped prove her innocence. But as she herself pointed out to the media, this mistake affected her health, damaged her reputation and caused her psychological stress.
Although it may seem to be an isolated event, the truth is that it is not a particularly unusual case, even with the best ML training and the highest accuracy rate, AIs are not infallible. In this particular case, the “hallucination” may have resulted from a simple confusion in establishing an acceptable similarity between the suspect and Porcha, according to the reference values.
See also: How to bring value to my company through Data Analytics?
Although the legal framework is no stranger to this type of confusion, since similar cases have been reported when there are eyewitnesses for various reasons, it is striking that there are no clearly defined protocols for those affected to defend themselves in this type of failure. It is also noteworthy the complete trust given to AI, allowing more and more autonomy without adequate supervision.
Beyond the failures, this generates an alarming concern, when you consider the malicious uses of it, and that is that legal loopholes that do not consider the use of AI, are a major security breach for any legal system. We simply cannot assume that everyone will make responsible and ethical use of this technology, so proper regulation of its implementation is vital.
Lost Cause?, AI in the industry
At the center of the debate, uncertainty is growing and while there is fierce resistance to implementation in different guilds, the facts cannot be denied according to the IBM Global AI Adoption Index 2022 study, conducted with Morning Consult and in which 7,500 companies collaborated, the global AI adoption rate is 35%.
In addition, according to the consulting firm McKinsey, this technology could generate $2.6 billion dollars per year. A percentage of implementation is still very low, however, these first glimpses of change are impressive and have kept CIOs, businesses and organizations on the lookout for the conversation and evolution of AI.
Intrigue, excitement and concern are everywhere when AI becomes the center of the conversation. The challenge is to be able to efficiently integrate this technology into businesses and to understand how to generate value in each business without being rushed, in order to avoid falling prey to mistakes.
Keep exploring: Science, Data, Cloud and Rock & Roll.
Currently, most of the companies that are already applying AI are using it to automate IT-related processes, improve the performance of their networks and communications, reduce costs, find optimization points and improve customer experience, according to the IBM study.
In this sense, it is pertinent to remember that the scope of AI is not limited to the administrative part and can also be implemented in the operational part, so refusing to accept the adoption of this technology is a lost cause, however, as has been emphasized, this integration must be in a responsible and ethical manner.
With this in mind, it is important to consider how to avoid job displacement, what are the barriers to adoption, and how to get the greatest return from AI?
Practical uses and responsible integration
At this point, the uncertainty that arises with every small step can be overwhelming, which is why it is necessary to take a breath and take a look at the big picture, remembering that the key aspect is “integration”. And what better way to do this than by reviewing some of the most recurring applications of AI in Industry.
- Technological democratization
One of the greatest fears regarding AI and other technologies is the level of expertise required to operate these technologies; however, AI, particularly generative AI, is very useful in bridging this digital divide, thanks to its understanding of natural language, it is capable of receiving and interpreting information in a simple way, providing equally understandable answers.
This, in addition to helping to simplify their integration, avoiding dependence on IT specialists to operate new technologies, also contributes to reducing costs, which is often another barrier to the adoption of new technologies. - Digital Debt
Hand in hand with the previous point, another frequent use of AI is to help pay off the digital debt, that is, to help employees and collaborators not to be overwhelmed by the volume of digital interactions on fronts such as emails, chats, notifications or information processing. This translates into more efficient responses, better performance in creative work, and therefore better communication in the organization. - Consumer experience
Another aspect where AI stands out is raising the level of consumer satisfaction, from multiple fronts, such as chatbots to solve frequent problems, follow up on customer service, evaluation of after-sales service, to aspects related to the supply chain, such as strategic suggestions to develop data driven strategies, or alerts on operations of interest to develop new services. - Data Analysis
Equally related, one of the most important uses of AI is that of data analytics, from which multiple solutions can be derived, since in addition to identifying consumer preferences, it also allows us to evaluate the behavior of our collaborators and even competitors. Key aspects in the development of data driven strategies.
When they are taken advantage of in a timely manner, in addition to great reductions in operating costs, they provide a greater return on investment and new business opportunities, allowing the company to take advantage of the differentiating factors of the competition.
It also allows us to evaluate the performance of multiple areas of operation and establish optimization objectives with precise follow-up metrics, which offers multiple advantages by providing an objective evaluation of our processes, allowing us to identify areas of opportunity and strengthening.
- Software Development
Have you heard that technology improves itself; there is something true, and the field of programming is one of the fields that has taken best advantage of the development of AI, since by incorporating it as a co-pilot of the developer, it allows a more intuitive and comfortable adoption of the DevelOps philosophy, from which others such as MLOps and APIOps derive.
They encourage more efficient software development by complying with standardized protocols, which allows testing, optimizations and documentation to be carried out in an accelerated and efficient manner, thus contributing to increase programmers’ productivity. In addition to allowing a better understanding for inexperienced users, as mentioned in a previous point. - Security
We have already talked about its use for facial recognition, however the security sector is one of the sectors that has explored the use of AI the most. An example is the 24/7 surveillance from sensors and cameras to detect unusual activity patterns and report from notifications, suspicious activity, without being limited to criminal activity, as it can also detect other dangers such as fires, etc.
Another application is the use of automated security dogs in the private sector, but of course we cannot leave aside its applications in cybersecurity, where it has benefited multifactor authentication, fraud detection and prevention and more.
A particular case is what has been achieved with Google’s Security Command Center, which is capable of neutralizing threats in real time to safeguard data privacy, in addition to offering security recommendations and being able to monitor entire cloud environments. This not only translates into a significant reduction in cybersecurity costs, but also prevents potential losses and provides flexibility in work environments. - Resource distribution
The food and economic industries are two sectors that have taken advantage of AI to efficiently redistribute their resources, in the case of the financial sector, the adoption of Open Banking has put traditional financial institutions out of business, bringing banking to the customers and not the other way around.
This has translated into an increase in demand from banking clients, with Fintechs being the best example of this. On the other hand, the food sector has benefited from the integration with delivery services that expand its potential reach, and digital attention that reduces the endless lines to order, to the inventory chain, reducing food waste by up to 40%.
In addition, an improvement in the supply chain is achieved by optimizing delivery routes, which not only translates into faster deliveries, but also into a sustainable improvement, since by reducing delivery times, the emission of polluting gases and fuel consumption are also reduced.
- Content Creation
At the beginning of this article we briefly mentioned how the Hollywood strike is evidence of the need to regulate this technology. However, this is also evidence of the applications of AI in content creation and, as mentioned earlier, generative AIs have put the spotlight on this technology.
AI is capable of creating royalty-free images, audio and writing, mostly used in the marketing and entertainment industries.
With a clearer perspective of some applications of this technology, let’s talk about how to responsibly integrate this technology, for this it is key to understand that AI is able to automate multiple tasks and processes, however it is not a substitute for the worker.
Without wishing to enter into further controversy regarding the debate on its scope, we must understand that although AI is evolving at an accelerated pace, we are still understanding how to get the most out of it. We are in an early adoption phase, which means that we must monitor its proper functioning on an ongoing basis.
In this sense, despite its capacity for autonomy and automation, it is better to maintain an approach to AI as a co-pilot of our workers, helping to maximize their performance and reduce operating costs, rather than as a definitive substitute.
To give a practical example, taking content creation, most marketers have relied on generative AIs such as Bard to write copy, find inspiration, or broaden their approach, using the responses from these AIs as a basis to polish them using their expertise to arrive at a final result by making the necessary adjustments.
In other words, although you improve the efficiency of the creatives and can increase performance, the human factor does not disappear, and understanding this is essential to understanding responsible integration, as well as possible personnel reassignments.
Improvements in performance within your organization allow for greater operational flexibility, allowing you to consider factors such as emotional pay, expanding your business with controlled expenses and exploring new business models, improving employee satisfaction, which can translate into increased employee engagement, loyalty and productivity. - Maximizing profits
You now have a panoramic view of AI, and you will have realized that although the current level of adoption is at a lower rate, AI is already part of our daily lives, and there is a great deal of interest in incorporating it into more and more industry sectors.
I would dare to say that you are facing the beginning of a new industrial revolution, the one that arises from the digital transformation, so the phrase renew or die, has never been more accurate, but how to do it with the greatest benefits?
If you have been paying attention, AI is increasingly designed to require fewer experts to use and implement, however, you will also remember the importance of ML, with respect to AI efficiency.
This is where the experts come in, and of course, we know that you probably don’t plan to expand your company’s IT area, and if you’re in management, you may not have the time to become an AI expert. Fortunately, you have the help of Amarello, and the Google Cloud services catalog, thanks to which you can access solutions like Vertex AI.
Tool, thanks to which you will be able to access multiple AI solutions, develop your own, and establish supervised training for your AI, in an efficient and accessible way, accessing the relevant resources with a controlled investment.
But of course, every business is different, so we invite you to contact us to receive personalized advice and discover how to maximize profits in your organization.
