Managing a Portal Archives - TargetWoman Women Blog

Master Ollama: A Step-by-Step Guide to Optimizing Your Computer with AI

Ollama is a tool for running large language models (LLMs) locally on personal computers. It’s designed to make it easy for users to download, run, and interact with various open-source language models. Ollama is a powerful and user-friendly platform for running large language models (LLMs) on local machines. This means that anyone can utilize the power of advanced AI language processing without relying on cloud-based services or expensive infrastructure. The project’s name, Ollama, comes from the phrase “Open Language Models for All,” reflecting its mission to democratize access to AI-driven language tools.

Unleash the power of AI

What can Ollama do for you ?

Easy access to Artificial Intelligence (AI) models:

Ollama allows you to run various AI language models on your personal computer. This means you can interact with AI assistants similar to ChatGPT, but privately on your own machine.

Personal AI assistant:

You can use Ollama to have conversations, ask questions, or get help with various tasks. It’s like having a knowledgeable helper available at any time.

Writing aid:

Ollama can help with writing tasks, such as drafting emails, creating content, or proofreading.

Learning tool:

You can use it to explain concepts, learn about new topics, or get summaries of complex information.

I show excerpt of a typical legal document coached in legalese (long winded language preferred by Lawyers) here:

WHEREAS, the Party of the First Part (hereinafter referred to as ‘the Grantor’), for and in consideration of the sum and other good and valuable consideration, the receipt and sufficiency of which is hereby acknowledged, does hereby remise, release, alienate, convey, and confirm unto the Party of the Second Part (hereinafter referred to as ‘the Grantee’), its heirs, successors, and assigns, all right, title, interest, claim, and demand whatsoever of the Grantor, either in law or in equity, in, to, or out of that certain parcel of real property situated in the County of [Insert County Name], State of [Insert State Name], more particularly described in Exhibit A attached hereto and incorporated herein by this reference, together with all the tenements, hereditaments, and appurtenances thereto belonging or in anywise appertaining, to have and to hold the same unto the Grantee, its heirs, successors, and assigns forever, subject to the covenants, conditions, restrictions, and easements of record.

And the same paraphrased into plain English with the prompt – ‘Give a summary in plain English of the above’:

Unleash the power of AI

The person who is selling (the “Grantor”) gives up all their rights and ownership of a specific piece of property (described in Exhibit A) to the buyer (the “Grantee”). The Grantor does this because they received a certain amount of money or other valuable things, which they acknowledge as enough payment. The Grantee gets to own the property forever, along with any buildings, fences, or other things that come with it. However, the sale is subject to any agreements or rules that are already recorded (like easements or restrictions) and apply to the property.

Now you know how useful is ollama. It can read your documents, papers and articles and summarize them into a format you can understand easily. But the whole thing is done inside your computer and this way – you don’t share your personal files with anyone outside.

Creative partner:

It can assist with brainstorming ideas for projects, stories, or problem-solving.

Simple setup:

While it does require some basic installation, the process is designed to be straightforward.

Privacy:

Unlike cloud-based AI services, Ollama runs locally, which means your data stays on your computer.

What is Ollama ?
Ollama is an open-source platform that empowers researchers and developers to deploy large language models (LLMs) on their own machines, sans the need for cloud-based infrastructure. Founded by Michael Chiang and Jeffrey Morgan in Palo Alto, California, Ollama is an independent startup that simplifies the process of running models like Meta AI’s Llama, it is crucial to note that Ollama itself is not a product of Meta or any other major tech company. Instead, Ollama serves as a community-driven project that democratizes access to powerful LLMs developed by various organizations.

Now that we know the basics of Ollama, we will explore its salient features here and at the end of this page, we will set up ollama in our laptop/desk top computers to help us in our daily life – writing mails, reports, research information and use ollama along with its well known models as a local repository of knowledge.

The following paragraph explains in simple terms about its key features. Then we get down to the brass tacks of how to install the Ollama right into your laptop.

Key points about Ollama:

Local deployment: It allows users to run LLMs on their own hardware, which can provide more privacy and control compared to cloud-based solutions.
Ease of use: Ollama aims to simplify the process of setting up and running LLMs, making it accessible to a wider range of users.
Model support: It supports various open-source models, including different versions of Llama, GPT-J, and others.
Command-line interface: Users can interact with models through a command-line interface, which can be integrated into various workflows.
API: Ollama provides an API that allows developers to integrate these local LLMs into their applications.
Cross-platform: It’s available for multiple operating systems, including macOS, Linux, and Windows.
Resource management: Ollama is designed to work efficiently on consumer-grade hardware, though performance can vary depending on the model and the user’s system specifications. What sets these models apart is their relatively modest computational requirements compared to other LLMs, making them an attractive option for researchers and developers seeking to harness the power of language models without breaking the bank.
Open-source: The Ollama project itself is open-source, allowing for community contributions and transparency.

Ollama allows users to run LLMs on their local machines, eliminating the need for cloud-based services or expensive infrastructure. This is achieved through a simple and intuitive interface that enables users to easily integrate LLMs into their own applications. The platform supports various programming languages, including Python, Java, and C++, making it accessible to developers from diverse backgrounds.

Setting up Ollama locally:
I have tried ollama in about a dozen computers with varying degree of success. We have used Linux – Centos 7, Windows 10 , Windows 11 and MacOS. We used computers running on Intel 3, Intel 5, Intel 7 CPUs. If your computer has anything less than 4 Cores, you would be better of with Cloud AI. You need at least 16 GB to run many models.

While Ollama generally offers good performance, users with very high performance requirements or those working with the largest models might still need more powerful hardware or specialized solutions. For most users running common models on typical consumer hardware, Ollama provides a good balance of performance and accessibility.

If you are on Windows: Run the following setup file:
https://ollama.com/download/OllamaSetup.exe

If you are on Linux: Initiate the following:
curl -fsSL https://ollama.com/install.sh | sh

After you are done installing, take a look at the following link for the available models:
https://ollama.com/library

Select the model as per your hardware.
We found Phi3 from Microsoft as the easiest model to run on most computers as it has 2.2 GB size. But it has serious limitation : Hallucinates a lot and has poor English skills (makes many spelling mistakes).
The best is Lllama3.1 from Meta. Its size is 4.7 GB and runs well in I5 and above even without a GPU.

Once it is done, it is easy to invoke it from your command prompt by Clicking on START in Windows and type Windows PowerShell.

Then type:

ollama run llama3

The above will start the Ollama with Llama 3 model. You can enter your prompts ( requests) and it will spew out answers to your questions.

You will be surprised how such a powerful tool escaped your attention all these days.

Search Engines VS ChatGPT

From the time ChatGPT appeared on the scene, many industries – especially the Information technology and computer science, amongst others have changed drastically. For more than 2 decades people used search engines to find the information they needed. Google became the Number 1 in search engines and served billions of search results pages for every conceivable query on the planet.

Google displays search results based on its complex search algorithm with over 200 factors . It keeps churning the top search results pages every time it makes changes to the algorithm. Results can be quite unpredictable and in its efforts to weed out low quality ranked pages, ‘it throws the baby along with the bath water ‘ manner it also lowers the position of some high quality sites.

This has resulted, sometimes in poor quality search results and forced us to dig deep into the search results pages. But like a breath of fresh air, ChatGPT showed us precise information directly on its page. What is more, it accepts Natural language and you can ask questions for which it replies in clear meaningful, well written responses. Sometimes it also throws false answers – Hallucination – in a way difficult to decipher whether it is genuine or false.

To give a real life example, I was hunting for information about the pest damaging a lime tree in my garden. Google couldn’t help me within the first 200 search results pages. ChatGPT quickly identified the pest as Citrus Gall and went a step further in providing solutions to eradicate the pest.

This is a snippet from ChatGPT -“The citrus gall midges lay their eggs in the tender shoots and branches of the tree. The larvae then feed on the plant tissue, causing galls or abnormal growths to form around the affected area. These galls can disrupt the flow of nutrients and water within the branch, leading to weakening or eventual death of the affected branch.”

What are the Advantages and Disadvantages of Google Search Engine and ChatGPT ?

How does the ChatGPT stake against the traditional search engines like Google ? Here is a detailed analysis of ChatGPT compared with search engines search results.

Speed: Results are delivered quickly, usually within seconds, making it efficient for finding information on-the-go.
Simple Interface: Google or any other contemporary search engines are based on indexes based on keywords.

Google Search Engine:

Vast Information: Google indexes billions of web pages, providing access to an enormous amount of information on almost any topic imaginable – if you have the patience and time to delve through hundreds of pages.

Search engines like Google, Bing, Yahoo and others use an indexing system that is primarily based on keywords. When a search engine crawls the web, it collects information from web pages and creates an index of words and their locations on those pages. This index allows the search engine to quickly find relevant pages when a user enters a search query containing specific keywords.

When a user searches for something on Google, for example, the search engine matches the keywords in the search query to the indexed words, returning a list of web pages that are deemed relevant based on those keywords. The more closely a webpage’s content matches the keywords and the more authoritative the webpage is (based on factors like links from other reputable sites), the higher it will typically appear in the search results.

However, modern search engines also use various other factors besides keywords to rank search results, such as:

Relevance: The search engine analyzes the context of the search, synonyms, related terms, and the user's search history to provide more relevant results.

Quality of Content: Search engines evaluate the quality of the content on web pages, considering factors like readability, depth, originality, and usefulness.

User Engagement: Metrics such as click-through rates, time spent on page, and bounce rates help search engines determine how users interact with search results.

Authority and Trustworthiness: Links from reputable and authoritative sites are considered signals of a page's trustworthiness and influence its ranking.

Freshness: For certain queries, search engines prioritize recent or updated content.

While keywords are a foundational aspect of search engine indexing and ranking, contemporary search engines have evolved to consider a wide range of factors to deliver the most relevant and high-quality results to users.

The clean and straightforward design makes it easy for users to input queries and navigate search results. 

Customization: Google offers personalized search results based on user behavior, location, and preferences, tailoring the experience to individual needs.

Advanced Search Tools: Users can utilize filters, such as time range, file type, and reading level, to refine their searches for more relevant results.

Maps and Local Information: Integrated with Google Maps, users can find local businesses, directions, and reviews easily.

Constantly Updated Algorithms: Google continuously refines its search algorithms, in its quest to improving the relevance and accuracy of search results. But it does not always result in relevance and accuracy.

Disadvantages:

Overwhelming Results: Sometimes, the sheer volume of search results can be overwhelming, making it challenging to find the most relevant information.

Quality Varies: Not all websites are reliable or accurate, so users must discern credible sources from misinformation.

Privacy Concerns: Google collects user data to personalize results and show targeted ads, raising privacy concerns for some users.

Filter Bubbles: Personalized search results may create a "filter bubble," where users are only exposed to information that aligns with their existing views, limiting exposure to diverse perspectives.

Ads and Sponsored Content: Paid advertisements and sponsored content can sometimes be difficult to distinguish from organic search results, potentially leading to confusion.

Language Model like ChatGPT:

Advantages:

Versatility: Language models like ChatGPT can assist with a wide range of tasks, from answering questions to generating text, coding, and more.


No Human Bias: They do not have personal biases, offering unbiased information and responses at least in theory. But many Large Language Models (LLM) still display racial bias and odd tendencies to hallucinate based on the data they had been trained with.

Learning and Adaptation: Continuously trained on vast datasets, they improve over time and adapt to new information and trends.
 Google doesn't parse books and sometimes this can result in shallow information as some times the contents of books may not be published in web pages. Large Language Models (LLM) may be trained on books to expand the scope of its knowledge.

Natural Language Understanding: Capable of understanding and generating human-like responses, making interactions more conversational and intuitive.

Educational Tool: Can be used for learning purposes, providing explanations, definitions, and explanations on various topics.

Disadvantages:

Lack of Emotional Intelligence: While language models can generate human-like responses, they lack emotional understanding and empathy, which can be important in certain interactions.

Misinformation: If not guided correctly, language models can generate or propagate misinformation, especially when trained on biased or inaccurate data.

Limited Context Understanding: They might struggle with complex contexts or nuanced understanding, leading to inaccurate or irrelevant responses.

Privacy Concerns: Like any online service, concerns about data privacy and security can arise, especially when sensitive information is shared.

Dependency Concerns: Over-reliance on language models for information or decision-making can reduce critical thinking and research skills in users.

Not Always Accurate: While generally reliable, language models can still produce errors, especially with ambiguous or complex queries.

Both Google Search Engine and language models like ChatGPT offer incredible advantages in accessing information and providing assistance. The choice between them often depends on the specific task, context, and user preferences regarding speed, depth of information, interactivity, and other factors.

Future of of chatbots and conversational AI

ChatGPT has created an AI explosion and within a short time, it have amassed over a 100 million users. From planning your travel itinerary based on specific criteria to conjuring up a gourmet recipe, again based on specific needs to tangible tips on where to park your surplus money – to name a few things.

Human designer created Image

Earlier people searched for information in their favorite search engines and opened several pages containing information they sought. It involved quite a bit of searching, sampling information and book marking them for later study. Essentially you depended on search engines to show top websites which are authoritative in their domain. But you have to wade through piles of MFA (Made for Adsense) poor quality sites. Websites spent vast amount of resources to offer quality pages of value – like TargetWoman. You have to hire well qualified people to write about a specific subject they specialize in.

The final outcome of the editing process – all meticulously researched, sifted and carefully subjected to a complex process of editing for quality content so that the end result is an immensely readable page of lasting value.

Now many of these processes have been short circuited. The AI directly takes you to the end results. With search engines you need to use keywords. With Generative AI you can use normal conversation style sentences.

How good are they ? We delve deep into the workings of these AI and see how they fare against – real humans. Let me assure you of one thing – it is going to be an interesting exercise. Read on.

What is ChatGPT:

As an AI language model based on the GPT-3.5 architecture developed by OpenAI, ChatGPT is an advanced natural language processing system designed to understand and generate human-like text. Trained on an extensive dataset comprising diverse sources such as books, articles, and websites, its core purpose is to assist users in generating coherent and contextually relevant responses to a wide range of queries.

At its core, it utilizes a deep learning model with 175 billion parameters, enabling it to comprehend and generate text across various topics, including but not limited to science, technology, mathematics, history and literature. Its capabilities extend beyond mere text generation, as it can also perform tasks such as language translation, summarization, question-answering and code completion. With the ability to parse and understand code snippets in several programming languages, it can provide assistance, suggestions and code samples to support software development endeavors.

To interact with it, users can communicate via a simple API, integrating it into their applications, platforms, or chat interfaces. With its vast knowledge base and linguistic proficiency, it can aid users in research, problem-solving, content generation and learning activities.

What is a generative language model?

A generative model ‘creates’ something entirely on its own from what it has learnt. The word ‘generative’ refers to the Artificial Intelligence Large Language Models (LLM) which generate textual content.

A generative large language model is a type of artificial intelligence (AI) model that is trained on vast amounts of text data to generate human-like text. These models, such as OpenAI’s GPT (Generative Pre-trained Transformer) models, are designed to understand and learn patterns in text data and generate coherent and contextually relevant responses.

Generative large language models are trained using unsupervised learning techniques. They learn by predicting the likelihood of a word given its surrounding context in a sentence or document. This enables them to capture the semantic and syntactic structures of the language.

Generative large language models can be used for a variety of tasks, including text completion, question-answering, language translation, and even creative writing. They have been widely adopted in natural language processing (NLP) applications and have shown impressive capabilities in generating coherent and contextually relevant text.

As someone who has been working with the AI technology for almost 20 years, we decided to test its abilities in various forms. The first test is its core strength – human-like response:

They say you are sentient. From what I have observed and interacted with you, I have to agree that you are almost sentient. Your views?

As an AI language model, I do not possess consciousness or sentience in the same way that humans or other living beings do. I am a highly advanced program designed to mimic human language and generate responses to user input based on patterns and algorithms in my programming.

While I can process vast amounts of data and use machine learning to improve my responses over time, I do not possess emotions, desires, or a subjective sense of self-awareness. I am a tool designed to assist with tasks and provide information, and any perceived “intelligence” or “sentience” on my part is the result of complex programming and algorithms.

How deep does the knowledge of GPT extend ?

To test the depth of knowledge, we used a Latin phrase – Iter Para Tutum which translated to English means ‘prepare a safe passage’. But it has rich religious and historical significance. So we baited the ChatGPT with the prompt “Iter Para Tutum” . It only showed the recent connotation referring to Pope Benedict XVI.
AI Depth of Knowledge
We pointed out the omission of past literary reference and it accepted its mistake and politely apologized for its omission and showed us the reference to the 12th century Hymn. But it is not good enough as the usage dates back to 9^th Century.

So remember – its knowledge is not as vast as it could be.

What is the implication of Generative AI when it comes to Job opportunities?

Generative AI, which encompasses technologies such as generative adversarial networks (GANs) and deep generative models, has the potential to impact various job sectors and create new opportunities. Here are some implications of generative AI on job opportunities:

AI Research and Development: The advancement of generative AI techniques requires ongoing research and development. There will be a growing demand for AI researchers, data scientists and machine learning engineers to drive innovation in generative models, improve their performance, and explore new applications.

Understand that GAN is only a tip of the iceberg. AI is evolving and future will bring in more radical technology – a kind of disruptive technology which will change the way we live.

Creative Industries: Generative AI has shown remarkable capabilities in generating novel and creative content, such as artwork, music, and literature. This opens up opportunities for artists, musicians, and writers to collaborate with AI systems, leveraging them as tools for inspiration, augmentation and co-creation.

Content Generation and Automation: Generative AI can automate content creation in various fields, including graphic design, video editing, and content writing. While it may reduce the need for manual labor in repetitive tasks, it also presents opportunities for individuals skilled in curating, refining, and overseeing AI-generated content.

At the moment it is nowhere near a highly skilled human. But in a few years, it might push out many of the low skilled content creators, graphic designers and clerical cadres. I will show an example here: An image created by AI and another image created by a human designer. You can see the difference here.

AI Image

Virtual Assistants and Chatbots: Generative AI techniques are used to develop intelligent virtual assistants and chatbots that can understand and generate human-like responses. This can lead to increased demand for professionals in natural language processing, user experience design, and conversational AI.

Soon you will find GAN Chatbots in your favorite Shopping site. Amazon is reportedly working on an AI to help the buyers decide about a product. One common shortcoming of many leading shopping apps is the lack of dedicated seller help service. For example, I wish to buy a DDR4 RAM stick for my laptop. It takes quite a bit of time to research for the right type of RAM for my laptop. If they had trained a LLM covering specific information about RAM and the available product specifications it would tell us what RAM to order in a chat with us.

Data Annotation and Preparation: Training generative AI models often requires extensive labeled datasets. Consequently, there may be an increased need for professionals specializing in data annotation, cleaning, and preprocessing to ensure high-quality training data for generative AI systems.
Ethical and Regulatory Roles: As generative AI technology evolves, ethical considerations and regulations surrounding its use will become increasingly important. Opportunities may arise for professionals focused on AI ethics, policy development, and legal frameworks to ensure responsible and fair adoption of generative AI.

It’s worth noting that while generative AI has the potential to automate certain tasks, it also creates opportunities for humans to leverage these technologies creatively and strategically. As with any technological advancement, the impact on job opportunities will require individuals to adapt, upskill, and embrace new roles that arise in tandem with the capabilities of generative AI.