Since the ChatGPT breakthrough in early 2023, a number of different generative AIs have been brought onto the market – such as Google Gemini or DALL-E. We’ll tell you the advantages and disadvantages.
- With its two models, ChatGPT is currently the most widely used AI tool for generating content.
- The biggest competition comes from Google Gemini – this AI can, among other things, respond in different styles.
- Dall-E and Leonardo.AI are of great importance for generating images and drawings.
It is probably true that no technological development will bring greater progress in the next few years than artificial intelligence (AI). It is already being used by millions of people around the world – not least thanks to the big ChatGPT boom at the beginning of 2023. There are now dozens of different AI tools with different features – you can get an overview of them in our comparison.
What are generative AIs?
Generative AIs are applications and programs that are specialized in generating content from simple or complex text or visual input (so-called prompts) and in this way responding to the input. The best-known and most widespread example of generative AI are chatbots like ChatGPT. You can talk to them as if the AIs were real people.
But there are other types of generative AI tools: In addition to chatbots, a large area is the creation of visual content. Tools like Dall-E or Leonardo.AI can “draw” images or generate photos in seconds based on prompts – often even in different styles (such as cartoon, realistic or digital art). Advanced AI models like Sora, recently introduced by OpenAI, can even generate amazingly real videos.
Generative AIs in comparison
Below we will introduce you to the four most important generative AIs with their advantages and disadvantages and tell you what you need to know about the range of functions and financing.
ChatGPT with GPT 3.5 and GPT-4: Advanced chatbot AI
ChatGPT is an AI chatbot developed by OpenAI, the free version of which is based on the GPT-3.5 AI model. ChatGPT is able to provide a suitable answer after the user enters text. The questions can be on any topic as long as they comply with the OpenAI Terms of Service, and the chatbot can respond in any manner and style desired by the user. The data basis for the chatbot lasts until the beginning of 2022 – so the AI cannot give you any information about current topics.
The AI is not connected to the Internet and therefore cannot research sources. Like many other chat AIs, it also tends to freely invent information, especially when it comes to scientific topics – so it is also not suitable as a source. However, thanks to the huge amount of data used as a basis, ChatGPT’s knowledge is still extremely extensive.
In the Plus version, which costs $20 per month (plus taxes), you also get access to the more advanced and modern GPT-4 model, which can answer even more complex questions precisely and work more creatively. ChatGPT Plus also provides access to DALL-E 3 – more on that below.
Google Gemini: In-house model with Ultra option
Do you remember Google Bard, the rather unpromising AI from Google? After a rebranding to Gemini, the chatbot was massively expanded and is now even ahead of ChatGPT in terms of functionality.
Particularly impressive: results generated by Gemini can be checked for truth with just one click and supported with suitable sources from Google searches. This way it is more largely guaranteed that the AI won’t tell you any nonsense. This feature (and many others, such as photo analysis) is even available in the free version.
Anyone who opts for a Google One AI Premium subscription for 22 dollars per month also gets access to Gemini Advanced with the 1.0 Ultra model, which can provide even more precise and creative answers – many experts even consider this model to be better than GPT-4 . However, the answers are primarily optimized in English; there could be minor inaccuracies in the translations into German.
DALL-E 2 and 3: Image generation at its best
As I said, in addition to chatbots like ChatGPT, there are also AI tools that are particularly concerned with generating photos and images. The DALL-E model, also developed by OpenAI, is exactly that: after a short text input (e.g. “Oil painting in the style of Monet, showing the Reichstag building on a sunny summer day”) you get four generated images which you can choose. In the free version you receive 15 credits, which can be used to generate four photos of DALL-E 2.
This is what the Monet Reichstag building generated by DALL-E 2 looks like. (Source: Generated by Netzwelt with DALL-E 2)
Anyone who pays $20 for ChatGPT Plus also gets access to DALL-E 3 in the same interface – the much more advanced model, which is also much more successful at generating text on images. The images can be generated with far more nuance and detail. You can also brainstorm content together with ChatGPT using GPT-4 and continually ask for improvements or additional content on already generated photos.
Leonardo.AI: Stable Diffusion-based image AI with editing capabilities
While DALL-E is a non-open source (i.e. proprietary) AI model developed specifically by OpenAI, Leonardo.AI is based on the open source stable diffusion model. This ensures that the content generated by Leonardo.AI is more predictable. The company also implements various helpful tools in its AI suite with which you can post-process photos that have already been generated using AI. In the free version there is a daily quota of 150 credits – one generation costs you 20 credits, post-processing costs correspondingly less.
We commissioned Leonardo.AI with the same prompt as with DALL-E 2 – this is where a weakness in the model becomes visible. (Source: Generated by Netzwelt with Leonardo.AI)
However, the model shows some weaknesses compared to DALL-E. The content is sometimes less relevant to reality than in DALL-E. This can be clearly seen in the example above: The green meadow, i.e. the Republic Square in front of the Reichstag, was replaced by a pond like a castle park. In return, AI can be used all the better for creative concepts developed from scratch. There are three levels of the payment plan – for 12 dollars per month you get 8,500 credits per month and access to additional features.
- » Tipp: The best VPN providers for more security and data protection
- » Buy balcony power plant: Comparison of the best solar systems
Don’t miss anything with this NETWORK WORLDNewsletter
Every Friday: The most informative and entertaining summary from the world of technology!
Table of Contents
