Generative AI models can generate original texts, images, videos, and code from existing content. And as you already know, they have colossal potential in life and at work.
GPT-4 has more capabilities than its older brother and performs the old ones better. It can provide more nuanced responses, can cite sources, handle intricate problems, analyze images. It’s generally more creative and less likely to “hallucinate.” And it only keeps improving. Midjourney also had a big boost recently. Compared to its previous version, V5 produces crazily detailed faces and limbs, supports a higher resolution, and produces generally more photorealistic images.
Both ChatGPT and Midjourney are known as generative AI models, meaning that they can generate original texts, images, videos, and code from existing content. And as you already know, they have colossal potential in life and at work. Here’s what generative AI can do for you and how it’s already being used in business.
Traditionally, such models work using Generative Adversarial Networks or GANs. In this sophisticated approach, two neural networks are pitted against each other in a zero-sum game. One neural net – a generator – is trained to create realistic input, while the other network – a discriminator – has to decide whether this input is fake or actually real from the training data. When a generator is finally capable of creating a fake input that can fool a discriminator, a GANs model is considered a success. That’s exactly how Midjourney and Jasper Art create such photorealistic images. GANs were introduced back in 2014, and in 2017 a different generative AI model was invented at Google. These neural networks are called transformers.
To put it simply, transformers look at the context around items – for example, words in a sentence – and predict what the next item or a word might be. They have what’s called an attention mechanism which allows them to detect intricate connections and dependencies between data elements.
One of the models built on transformers is LaMDA or Language Model for Dialogue Applications developed by Google. It’s so convincing at dialogue tasks that in 2022, Google’s engineer Blake Lemoine shared with the world that he believed LaMDA was sentient. Fascinated with some of AI’s responses, Lemoine was under the impression that the model has a soul and is afraid of death. He was subsequently fired for breaching the confidentiality agreement.
All in all, the scientific community is confident that we don’t have the technology to produce truly sentient robots and LaMDA is just that good at mimicking human speech. Another prominent example of generative AI using transformers is, of course, Generative Pre-trained Transformer or GPT, developed by OpenAI. Until recently, GPT-3 was the most complex model capable of producing human-like speech that could make virtually anyone believe that it was a real person talking.
In 2022, OpenAI launched a spin-off ChatGPT, modified specifically for conversational tasks, which captured the world’s attention and gave each of us a tireless assistant to help in personal and business matters. It can craft an evocative email, write a piece of code, create a marketing content plan, help with your research, and do many of the things that we used to do ourselves just a few months ago. Recently, Google unveiled Bard – a conversational AI originally powered by LaMDA with features comparable to those of ChatGPT. Google keeps opening up Bard to more countries and recently moved it to their more advanced model called PaLM 2. Most likely, it will soon become another handy tool for those who enjoy ChatGPT’s capabilities.
Here’s what generative AI can do for you and how it’s already being used in business
Use case 1 : Content Creation and Marketing Automation
![Content Creation and Marketing Automation](data:image/svg+xml;nitro-empty-id=MTAzNDo4MjA=-1;base64,PHN2ZyB2aWV3Qm94PSIwIDAgMTAyNCA1NDgiIHdpZHRoPSIxMDI0IiBoZWlnaHQ9IjU0OCIgeG1sbnM9Imh0dHA6Ly93d3cudzMub3JnLzIwMDAvc3ZnIj48L3N2Zz4=)
Generative AI can significantly enhance content creation and marketing efforts. While marketing itself is a creative and strategic task that you can’t fully trust AI to perform, there are many smaller tasks within it that can be easily automated. For example, crafting social media messages, sales copy, promotional text, and media is crucial for business recognition and success but is time-consuming and must be done almost daily. Special software like copy.ai and Postello can create engaging LinkedIn posts, Instagram captions, and YouTube video descriptions from a brief explanation. You can also use standard ChatGPT features to generate short articles, product descriptions, and other types of content. AI can be prompted to use a specific tone of voice or target a particular audience.
Whenever you need to accompany your text with images, AI will also come in handy. Often, it doesn’t make sense to engage a designer or illustrator and spend a whole day crafting a perfect image, especially if you’re a small business. You can simply describe what you want to tools like Jasper Art, Midjourney, or the AI tool in Canva, and get quick and cheap results. Depending on your experience with AI prompting, the final image can be used even in ads and promo materials online. AI can create original images and edit them, such as changing the background, removing unwanted objects, or replacing them with something else. In Canva, for example, you can brush over the area and type what you want to be added there.
If you’ve run out of ideas for blogs or don’t have time to research what content your customers engage with the most, AI can also help. It will provide you with a list of newsletter topics, ideas for marketing campaigns, and, if you choose to write your content personally, recommendations on the best way to do it. AI can also edit the text, change the tone of voice, remove redundancy, and help you perfect what’s already been written.
Use case 2: Concept Art and Idea Sharing
AI can assist in concept art and idea sharing. Although not sophisticated enough to replace traditional designers and illustrators for high-stakes tasks, AI can help you collaborate with them more effectively. When commissioning a designer, use generative AI to provide references that speak louder than words. Easily exchange ideas about the mood and aesthetic of the final design, character poses, lighting, and desired art style to understand each other better and quickly settle on the final design.
AI can also create storyboards. In video production, storyboarding helps visualize the script and plan scenes, though this task is often skipped due to its time-consuming and costly nature. Tools like Krock.io have introduced AI options for storyboarding. You can upload your text and have storyboards created automatically, streamlining video production. Since storyboards are just sketches that don’t make it into the final product, they don’t have to be perfect but can help prepare video content quicker, easier, and with better results.
Use case 3: Transforming Customer Service with AI Chatbots
![Transforming Customer Service with AI Chatbots](data:image/svg+xml;nitro-empty-id=MTA0MjoxMjE3-1;base64,PHN2ZyB2aWV3Qm94PSIwIDAgMTAyNCA1NjAiIHdpZHRoPSIxMDI0IiBoZWlnaHQ9IjU2MCIgeG1sbnM9Imh0dHA6Ly93d3cudzMub3JnLzIwMDAvc3ZnIj48L3N2Zz4=)
Generative AI can transform customer service. For many years, chatbots — virtual assistants meant to replace real company-client interactions — struggled to succeed. They were more of a marketing gimmick and couldn’t meaningfully converse with customers or provide real help. This is changing. Now, small startups and big corporations can use APIs from Google and OpenAI to integrate advanced language models and complement their capabilities. You can train your conversational AI to answer customers’ questions, retrieve requested data, offer guidance, or help with pain points. For example, when using a patient portal, a person can ask AI to schedule an appointment, remind them to take medication, or provide treatment information and health tips.
Additionally, you can create plugins for ChatGPT, allowing users to access information from third-party businesses within ChatGPT’s interface. Examples include plugins by Expedia and Kayak for travel recommendations, Instacart for orders, and OpenTable for booking places to eat. Such technology isn’t meant to replace human contact altogether but provides another option for interacting with your company, resulting in a competitive advantage and increased customer engagement.
Use case 4: Software Development
One of the biggest areas of applications for generative AI is in software development. In 2022, GitHub launched GitHub Copilot, built with OpenAI and trained on publicly available source code. This tool gives developers code suggestions in dozens of programming languages, matching the project’s context and style conventions. GitHub reported that with the tool’s assistance, developers were more productive, finished tasks significantly faster, and felt more fulfilled and satisfied.
A year later, GitHub made Copilot available for Business and introduced chat features, opening a variety of capabilities from code rewriting to adding comments. Integrating AI into engineering helps developers focus on finding solutions to complicated problems, which is crucial in a world where software is increasingly complex and economic pressures are significant. Although the implications of using AI in coding are not fully explored, it’s clear that it will become a paradigm shift comparable to the adoption of agile methodology.
Use case 5:Optimizing Operational Tasks
AI can help you optimize operational tasks. Any number of mundane tasks can be allocated to AI. For example, when you have a big document or research to go through, use AI to summarize it for your convenience. If you’re the one preparing the presentation, use generative tools to give you prompts on what to talk about or create images to illustrate your thoughts in slide decks. Instead of looking through myriads of stock photos, you can let your imagination run wild and simply describe what you’d like to see on the screen.
Use case 6: AI in HR Operations
There are also many applications for AI in HR operations. AI can assist in creating questions for candidate assessments or automate onboarding with AI-enabled internal portals, where employees can ask questions. Another example is a skill management platform TalentGuard that used OpenAI’s GPT model to create WorkforceGPT that helps with skills development in organizations.
Use case 7: Enhancing Ecommerce with AI
![Enhancing Ecommerce with AI](data:image/svg+xml;nitro-empty-id=MTA1Mzo1ODk=-1;base64,PHN2ZyB2aWV3Qm94PSIwIDAgOTI1IDc4NSIgd2lkdGg9IjkyNSIgaGVpZ2h0PSI3ODUiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyI+PC9zdmc+)
In ecommerce, AI can also be used to create realistic product photography. To make your product catalog more visually appealing, you can create multiple product photos without spending time and money on professional photoshoots. With Dream Booth, you can generate images of the product in different environments. Most generative AI tools work text-to-speech only, but this model, published by Google in 2022, can be retrained to incorporate your product and use it when generating images.
Generative AI Problems
![Generative AI Problems](data:image/svg+xml;nitro-empty-id=MTA1Nzo1NzM=-1;base64,PHN2ZyB2aWV3Qm94PSIwIDAgMTAyNCA2NTkiIHdpZHRoPSIxMDI0IiBoZWlnaHQ9IjY1OSIgeG1sbnM9Imh0dHA6Ly93d3cudzMub3JnLzIwMDAvc3ZnIj48L3N2Zz4=)
Intellectual Property Issues: Generative AI is trained on masses of publicly available data, including data protected by copyright. For end users, it’s virtually impossible to know what section of AI-generated content is under which license. So, what happens if the code produced by GitHub Copilot turns out to be copyrighted? Or when the art you’ve been using in your promo materials is eerily similar to someone else’s original art? There are no rules or regulations for using AI-generated data, which means there will always be risks. The good thing is that you, as a business, won’t likely be the one sued for using copyrighted content, but the company that built the model will.
Reliability Issues: Generative AI isn’t always reliable. It makes mistakes and can provide outwardly wrong results, which may be harmful or damaging to a business’s reputation. In any high-stakes scenario, you still need a person to check if the information returned by AI is factual and up to date. The lack of trust can also hinder adoption. If you don’t understand where the results are coming from, you can’t rely on them and thus, may choose to not use AI altogether. However, as AI develops and mistakes become rarer, this problem may disappear.
Recommendations for Organizations
1. Analyze Industry Impact: Start by analyzing how the technology might disrupt your industry and where it can introduce value. If this hasn’t happened yet, generative AI will impact the business, and it will happen soon. Gather a team of business leaders, the legal department, and other important specialists to predict the most popular use cases and think about how you can adapt to changes.
2. Develop a Generative AI Policy: Work on your generative AI policy. The team must understand when they’re welcome to use ChatGPT and when you expect them to perform daily tasks the traditional way. For example, if you create content about healthcare, perhaps you want it to be written by experts or at least reviewed by one before posting it.
3. Incorporate AI into Company Values: Don’t blindly adopt the technology but try to incorporate it into your company values and culture. This will keep distinguishing you from competitors, not the mere fact of using AI.
4. Foster Trust: Communicate your policy to your stakeholders and customers to make sure they know how and when you’re using AI and understand that behind every robot assistant, there are real people who continue bringing value.