AI tools are not just ChatGPT

1. ChatGPT is smaller than GPT-3
2. Artificial intelligence for all to-dos?
3. The commands matter
4. What can't AI do?!
5. The Wild West of Artificial Intelligence?
6. Get involved with AI as early as possible

It's probably already the topic of the year: Artificial Intelligence. Since the hype around ChatGPT, overwhelming excitement meets worries about jobs and fear of manipulation. But what can we expect in the future? This question was looked into by Jens Polomski, expert for Online Marketing, and Manuel Gerlach, Head of SEO and Content at OMR Reviews, in our online seminar. In addition to clarifying important terms - yes, there's no such thing as ChatGPT-3 - it naturally was also about AI tools and exciting use cases. You can read the recap here.

Missed our online seminar on AI tools? No problem: here you can access the recording afterwards.

ChatGPT is smaller than GPT-3

No later than the launch of OpenAI ChatGPT at the end of 2022, terms came up that have been around for a while. For many, however, they are new and often confused. And even though the majority of our participants had already worked with AI tools, they were completely new territory for a good 27 %. It helps to clarify the most important AI buzzwords in advance. Because: ChatGPT is often turned into ChatGPT-3 by connoisseurs - but it doesn't exist.

• OpenAIis the company behind the GPT language models (Generative Pretrained Transformer). These are language models that are trained using text data and are supposed to create content that sounds as human as possible. However, the GPT technology itself does not come from OpenAI. It was developed by Google in 2017.
• GPT-3 is the latest language model by OpenAI and is based on a chat interface. It was released in 2020 and is the successor to GPT-2 (2019). The technology consists of 175 billion parameters that emerged from training data such as plain and webtext and Wikipedia articles.
• ChatGPT is a chatbot based on GPT-3 and a refined model of this technology. It is optimized for conversation and consists of about 20 billion parameters - only a fraction of GPT-3. The tool was released on November 30, 2022.
• Many AI tools have been using the API interface to GPT-3 for a long time and thus directly work with the technology of OpenAI. With the launch of ChatGPT, OpenAI has now become a direct competitor to such language and text AI tools.

Artificial intelligence for all to-dos?

If you've ever tried ChatGPT or other AI tools, you can at least guess the possibilities they offer. In addition to pure research, they can be used for many tasks and thus can make your daily work easier.

Especially with regard to SEO and text creation, AI tools are already very strong. With them, you can e.g. query relevant keywords for your content. ChatGPT then even summarizes them for you, including additional informations such as explanations and search intent, in handy tables. Provided the command is correct. More about that later.

First, Manuel gets ChatGPT to create a topic list.

Then he asks for relevant keywords for each topic.

Finally, he gets the information summarized in a clear table.

In addition, text AI tools can help you create blog posts or give you an outline to start with. Particularly comprehensive here is e.g. neuroflash, which offers you different workflows. They are optimized for content like blog articles, LinkedIn posts or other social media formats. So the Hamburg start-up does not just use the API interface to GPT-3, but also develops helpful work processes for different formats. By the way, on OMR Reviews you can learn how neuroflash also wants to become the technology people love.

If you're dealing with SEO, you might be wondering how Google actually responds to AI content. And because AI tools mainly make your work easier and do not (yet) completely take it away, it depends on the quality and added value. In general, you should always check AI-generated content and give it a final polish. But then the tools really do a lot of tasks for you. What else besides product texts - keyword: Shopify-integration - and whether Jens already has a AI manage his emails? Stay tuned.

The commands matter

AI tools like ChatGPT are like a very smart (and very helpful!) colleague: You can ask them anything and they can do the most tedious to-dos for you. However, the important thing here is that you speak clearly. Because only clear instructions lead to clear results. So far, so logical.

These instructions or commands are also called prompts. To get the desired result, you should clearly define your goal and expectation. Information and examples are also helpful: For example, which topics should be particularly addressed or which questions absolutely need to be clarified? Things get really interesting when a tool completes a series of tasks: For example, it can adapt content or compile it, like above, in tables with the help of follow-up tasks.

Fortunately, you don't have to complete training as a Prompt Engineer to create effective prompts. Some relief is provided by tools like the Chrome Extension AIPRM for ChatGPT. With it, you can fall back on tried and tested prompts and share your own with the community. And our experts also have a helper for Google Docs and Sheets in their online seminar.

What can't AI do?!

AI-generated texts are almost old news after three months of ChatGPT. Looking into the future is particularly exciting: What else awaits us? One hot topic is AI images. The technology here is getting increasingly better. However, there is still plenty of room for improvement when creating images featuring humans, as Manuel's LinkedIn post shows:

Nevertheless, AI tools already offer many possibilities for image creation. What these are and what you should keep in mind with prompts, you can read in our article on AI images.

Video is also a big topic besides image creation. Until text-to-video runs smoothly - i.e. creating video content from text prompts - it will take a while. However, some AI tools can already edit videos. Video avatars are also a thing.

Text-to-audio and audio-to-text are a bit better off. And then there's audio-to-audio: something like voice cloning. At the very least at this point and also looking at the video avatars, a certain unease is definitely justified. Despite all the possibilities and cool use cases, AI and its rapid dissemination also entail risks. Helpers like in Microsoft Teams Premium or Whisper from OpenAI still offer a lot of potential for meetings or communication with customers.

The Wild West of Artificial Intelligence?

"Elon Musk sees AI as one of the greatest risks to humanity", headlined the digital magazine t3n just the other day. Necessary regulations are lagging behind the rapid spread of AI tools and their functions in many areas. In addition to basic moral and ethical questions, legal aspects are of course also interesting. If the AI creates its content from existing sources, who is the intellectual owner? And what about data protection? The fact is that clear facts are often still lacking. The classic gray area.

What you should also keep in mind besides legal questions: The data with which GPT-3 was trained comes from 2021. Everything that happened afterwards - and that's quite a lot - the technology didn't capture at all. So if you want to work on current topics in particular, you also need to take action yourself.

Get involved with AI as early as possible

The headlines of the last weeks and months arouse the feeling that artificial intelligence is a brand new topic. In fact, we have been surrounded by more AI than many people think in everyday life for years. For example, the translation tool DeepL is already working with a strong AI focus. And the e-bike brand Sushi Bikes has already implemented a whole advertising campaign using AI. But also more everyday things like email spam filters and face recognition rely on artificial intelligence. Nevertheless, not everything contains AI. Sometimes it can be just plain old automations.

And what about the future? First, rest easy: AI won't take over your job overnight. But that doesn't mean you should ignore it. The tip from our two experts is to get involved with AI tools as early as possible. You've already made a start here. Because at some point these tools will become commonplace. And then using ChatGPT shouldn't make you feel awkward anymore. Instead, you might be asked, “Why didn't you use ChatGPT for that task?” The best way to prepare for this is with our 10 tricks for ChatGPT.

By the way, maybe that's also the point when we don't care how much AI support Manuel got for promoting the online seminar and how Jens manages his emails. But it still matters. The solution as well as more tools and use cases can be found in the recording: