Search for a Prompt Engineer IV - Narendra Modi, Gemini, and Fascism

Hardware defects, whether the Boeing’s Problems with the 737 MAX that caused hundreds of deaths and a door panel that flew off; or the O-ring seal failure in the right Space Shuttle Solid Rocket Booster due to cold weather and wind shears that caused the Space Shuttle Challenger disaster, are terribly hard to fix.  

And exorbitant. 

Software defects, however, are cheap and easy, and most often do not have a direct human impact. All you need to do is debug, test, ship and update, and you are good to go. My iPhone updates while I am asleep, I have been told Tesla does the same. Even when it goes wrong, there is not much fallout, up until your AI Bot calls the Prime Minister of India, a fascist.  

Before we get to the toxic combination of inexperience, hubris or lack of cultural understanding that led to this mistake. Let's take a cursory look at the definition of AI. 

AI is software that does fancy data processing.

Basically, Artificial intelligence (AI) is the simulation of human intelligence processes by machines, especially computer systems, and to that it relies on Data.  

'And therein', as the Bard would say, 'lies the rub.' 

AI problems are Data problems; this also makes them human problems. And the government of India either wants to find out who was that human and either neutralize him or educate him, so he sees the light.  

But jokes apart, the ML bias shown by the AI was directed towards a poor country's Prime Minister and for similar queries Gemini did not display similar sentiments for leaders of the western world.  

It does not seem by design, it was simply ignorance on their part and probably the importance of the Indian market, which is of lesser value and as a result of lesser consideration.  

But Political Economy aside, Gemini shows bias, whereas Google Search provides straightforward results, and Google is answerable for the former and not so much for the latter.  

This is a bias in action for you. Mother Teresa fighting poverty. 

Which is more akin to Russel Crowe Fighting Cancer in South Park. 

Your AI Software is as good as your data and would probably work with a sub part language model but will not survive bad data, and its importance cannot be overstated. 

Data plays a crucial role in building an AI system, Data is the foundation of AI systems, shaping their capabilities, performance, and reliability. The collection, preparation, and maintenance of high-quality data are critical steps in building AI models that can effectively tackle real-world challenges. 

Here are several key aspects highlighting the significance of data in the development of AI systems: 

    Training Models: 

AI systems, especially machine learning models, require large amounts of labeled data for training. The quality and quantity of training data directly impact the performance of the model. The model learns patterns, features, and relationships from the input data during the training process, enabling it to make accurate predictions or decisions. 

    Generalization: 

Adequate and diverse data helps AI models generalize well to unseen examples. Generalization is crucial for an AI system to perform effectively on real-world data beyond the training set. 

    Algorithm Performance: 

The performance of AI algorithms is highly dependent on the quality of data they are trained on. Flawed or biased data can lead to biased models and inaccurate predictions. 

    Feature Extraction: 

Data provides the features or variables that the AI model uses to make predictions. The selection and representation of features play a significant role in the model's ability to understand and extract relevant information from the input. 

    Supervised Learning: 

In supervised learning, where the model is trained on labeled data, the ground truth labels guide the model to learn the mapping between input and output. The accuracy of these labels is critical for the model's performance. 

    Unsupervised Learning: 

Unsupervised learning models also rely on data, often without labeled examples, to discover patterns, clusters, or structures within the data. The quality and diversity of data influence the effectiveness of these algorithms. 

    Model Evaluation: 

Evaluating the performance of an AI model requires data for testing and validation. The model's ability to generalize and its robustness can be assessed using independent datasets. 

    Adaptability and Evolution: 

As AI systems operate in dynamic environments, the availability of real-time data is essential for continuous learning, adaptation, and evolution. This is particularly relevant in applications like reinforcement learning and online learning. 

    Handling Edge Cases: 

Robust AI systems need exposure to various scenarios and edge cases. Sufficient and representative data help ensure that the model can handle diverse inputs and unexpected situations. 

    Ethical Considerations: 

Biases present in the training data can be inherited by AI models, leading to biased outcomes. Ensuring fairness and mitigating biases requires careful curation and scrutiny of the training data. 

All of this is to say that ML bias will cause problems, in roughly the same kinds of ways as problems in the past, and will be resolvable and discoverable, or not, to roughly the same degree as they were in the past. Hence, the scenario for AI bias causing harm that is easiest to imagine is probably not one that comes from leading researchers at a major institution.  

Rather, it is a third tier technology contractor or software vendor that bolts together something out of open source components, libraries and tools that it doesn’t really understand and then sells it to an unsophisticated buyer that sees ‘AI’ on the sticker and doesn’t ask the right questions, gives it to minimum-wage employees and tells them to do whatever the ‘AI’ says. This is what happened with databases. This is not, particularly, an AI problem, or even a ‘software’ problem.  

It’s a ‘human’ problem.  

Artificial Intelligence at Hashbrown Systems 

Core of what we do involves working with ginormous datasets, and we have been in business for over a decade now and have learnt to work with Big Data in our own indigenous ways.  

The emergence of robust open-source models like Llama and 01.AI has empowered us to develop and implement AI software, revolutionizing user interactions within our platforms. This approach is not only remarkable but also cost-effective. Deploying AI solutions first, gathering feedback, and gradually introducing them has proven to be an effective strategy. 

What's next for us?  

Since we actively deploy AI tools into our daily processes and have been working with NLP and LLMs for quite a few years, expect to read and learn more on either Susan, our conversational AI for printing and converting machinery or Friday, recommendation bot for media allocation or both.  

Farewell, until our paths cross again.