What is Generative AI and How Can I Use It?
What is Generative AI and How Can I Use It?
What is Generative AI and How Can I Use It?
What does Generative AI (Gen-Ai) mean? What can I use Gen-Ai for? Which models are suitable for processing my concerns in a data protection-compliant manner? These and other questions will be addressed in the following sections and blog articles.
What does Generative AI (Gen-Ai) mean? What can I use Gen-Ai for? Which models are suitable for processing my concerns in a data protection-compliant manner? These and other questions will be addressed in the following sections and blog articles.
What does Generative AI (Gen-Ai) mean? What can I use Gen-Ai for? Which models are suitable for processing my concerns in a data protection-compliant manner? These and other questions will be addressed in the following sections and blog articles.
What does Generative AI (Gen-Ai) mean? What can I use Gen-Ai for? Which models are suitable for processing my concerns in a data protection-compliant manner? These and other questions will be addressed in the following sections and blog articles.
Understanding the basics of GenAi
The term Generative AI, often abbreviated as GenAi, describes software that can process and generate content in various forms. For example, in the form of text, images, speech, or music.
Models that can 'understand', process, and generate natural language are referred to as Large Language Models (LLMs). These models are typically trained on gigantic bodies of text. For instance, ChatGPT was trained on approximately 300 billion words. Most of these words come from scanned books and the internet, roughly equivalent to 4 million books.
LLMs can generate and process various forms of content, also called modalities, which can include the following:
Written text in 80+ different languages
Spoken text in 57+ languages
Numerical data
Programming code
Images
Videos
Music
GenAi is not an entirely new concept. Google Translate was founded in 2006. Siri, Google auto-complete, iPhone auto-complete. What has happened? Gpt-4 in 2023
Quality of outputs and hallucinations
The output of such systems is sometimes indistinguishable from texts written by humans. The quality of the output can be significantly influenced by prompt engineering. Additionally, various other factors influence the quality of the output, such as the amount and variety of data used to train the AI and the number of parameters (size) of a model.
However, GenAi systems are not always accurate and sometimes produce so-called hallucinations, which can be difficult to identify without context. Clever prompt engineering, retrieval augmentation, and fine-tuning are methods to reduce these hallucinations.
Commercial models vs. Open-Source M
Tools like ChatGPT, Gemini, Claude, and others are referred to as commercial LLMs. ChatGPT 4o is currently offered for free with a usage limit. Claude and Gemini offer so-called freemium models, allowing the AIs to be tested for free for a limited time.
The APIs of these providers are priced based on fees per input and output token, ranging between $5 and $75 per million generated tokens.
In addition to these, there are also open-source LLMs. These models have the significant advantage of being able to be installed and used on their own servers, thus bypassing data privacy concerns. Examples include: Llama 3, MPT-7B, Falcon 2, Vicuna-13B, Mixtral-8x22B, Grok, Command R
This technological innovation can be very profitable for businesses and will have a major impact on how content is effectively and efficiently created.
According to a McKinsey study [source], AI can contribute up to €4.4 trillion to the global economy each year. Within the next few years, telecommunications, media, and technologies without AI support will be considered obsolete or inefficient.
Understanding the basics of GenAi
The term Generative AI, often abbreviated as GenAi, describes software that can process and generate content in various forms. For example, in the form of text, images, speech, or music.
Models that can 'understand', process, and generate natural language are referred to as Large Language Models (LLMs). These models are typically trained on gigantic bodies of text. For instance, ChatGPT was trained on approximately 300 billion words. Most of these words come from scanned books and the internet, roughly equivalent to 4 million books.
LLMs can generate and process various forms of content, also called modalities, which can include the following:
Written text in 80+ different languages
Spoken text in 57+ languages
Numerical data
Programming code
Images
Videos
Music
GenAi is not an entirely new concept. Google Translate was founded in 2006. Siri, Google auto-complete, iPhone auto-complete. What has happened? Gpt-4 in 2023
Quality of outputs and hallucinations
The output of such systems is sometimes indistinguishable from texts written by humans. The quality of the output can be significantly influenced by prompt engineering. Additionally, various other factors influence the quality of the output, such as the amount and variety of data used to train the AI and the number of parameters (size) of a model.
However, GenAi systems are not always accurate and sometimes produce so-called hallucinations, which can be difficult to identify without context. Clever prompt engineering, retrieval augmentation, and fine-tuning are methods to reduce these hallucinations.
Commercial models vs. Open-Source M
Tools like ChatGPT, Gemini, Claude, and others are referred to as commercial LLMs. ChatGPT 4o is currently offered for free with a usage limit. Claude and Gemini offer so-called freemium models, allowing the AIs to be tested for free for a limited time.
The APIs of these providers are priced based on fees per input and output token, ranging between $5 and $75 per million generated tokens.
In addition to these, there are also open-source LLMs. These models have the significant advantage of being able to be installed and used on their own servers, thus bypassing data privacy concerns. Examples include: Llama 3, MPT-7B, Falcon 2, Vicuna-13B, Mixtral-8x22B, Grok, Command R
This technological innovation can be very profitable for businesses and will have a major impact on how content is effectively and efficiently created.
According to a McKinsey study [source], AI can contribute up to €4.4 trillion to the global economy each year. Within the next few years, telecommunications, media, and technologies without AI support will be considered obsolete or inefficient.
Understanding the basics of GenAi
The term Generative AI, often abbreviated as GenAi, describes software that can process and generate content in various forms. For example, in the form of text, images, speech, or music.
Models that can 'understand', process, and generate natural language are referred to as Large Language Models (LLMs). These models are typically trained on gigantic bodies of text. For instance, ChatGPT was trained on approximately 300 billion words. Most of these words come from scanned books and the internet, roughly equivalent to 4 million books.
LLMs can generate and process various forms of content, also called modalities, which can include the following:
Written text in 80+ different languages
Spoken text in 57+ languages
Numerical data
Programming code
Images
Videos
Music
GenAi is not an entirely new concept. Google Translate was founded in 2006. Siri, Google auto-complete, iPhone auto-complete. What has happened? Gpt-4 in 2023
Quality of outputs and hallucinations
The output of such systems is sometimes indistinguishable from texts written by humans. The quality of the output can be significantly influenced by prompt engineering. Additionally, various other factors influence the quality of the output, such as the amount and variety of data used to train the AI and the number of parameters (size) of a model.
However, GenAi systems are not always accurate and sometimes produce so-called hallucinations, which can be difficult to identify without context. Clever prompt engineering, retrieval augmentation, and fine-tuning are methods to reduce these hallucinations.
Commercial models vs. Open-Source M
Tools like ChatGPT, Gemini, Claude, and others are referred to as commercial LLMs. ChatGPT 4o is currently offered for free with a usage limit. Claude and Gemini offer so-called freemium models, allowing the AIs to be tested for free for a limited time.
The APIs of these providers are priced based on fees per input and output token, ranging between $5 and $75 per million generated tokens.
In addition to these, there are also open-source LLMs. These models have the significant advantage of being able to be installed and used on their own servers, thus bypassing data privacy concerns. Examples include: Llama 3, MPT-7B, Falcon 2, Vicuna-13B, Mixtral-8x22B, Grok, Command R
This technological innovation can be very profitable for businesses and will have a major impact on how content is effectively and efficiently created.
According to a McKinsey study [source], AI can contribute up to €4.4 trillion to the global economy each year. Within the next few years, telecommunications, media, and technologies without AI support will be considered obsolete or inefficient.
