Software development for connected devices and Internet of Things (IoT) applications are the talk of the town in today’s advanced technology era. It is not to forget that IoT projects involve intricate firmware development and includes managing of drivers, peripherals as well as device trees. IoT teams which aim for ambitious product goals have reason […]

Innovative startup EMoMee, working in the field of emotional intelligence (EQ) edutainment, has successfully secured $1 million in pre-seed funding co-led by prominent investors Whiteboard Capital, Gruhas Collective Consumer Fund and DeVC. The fresh capital signals strong confidence in the vision and potential of the company. EMoMee was founded by Pooja Jauhari, Varun Duggirala and […]

Raise Financial Services is in talks with investors to secure $100 million in funding. Fresh investments means the company’s valuation may reach to $1.2-1.5 billion. Currently it is valued at $150 million. It is the parent company behind Dhan trading app, which may step into the unicorn club. Dhan was founded by Pravin Jadhav, former […]

Could you provide to our TechiExpert readers an overview of String Metaverse and its mission in the virtual world? At String Metaverse, our mission is to revolutionize the virtual world by developing gaming platforms for a global audience of 7 billion people, including 3 billion gamers. We invest in Web 3.0 community tech platforms and […]

Are you an avid music lover or someone who likes to have offline access to your playlists wherever you are? A Spotify playlist downloader might be what you need. In this article, we recommend the 3 best Spotify playlist downloaders that you can use on the go, without cellular data or abroad. Let’s get started!  […]

In today’s competitive business landscape, exceptional customer service can set a company apart from its competitors. To achieve this, many companies are investing in advanced contact center software solutions that streamline operations, improve efficiency, and enhance the overall customer experience. Here, we explore several success stories of companies that have revolutionized their customer support with […]

On April 2, the World Health Organization launched a chatbot named SARAH to raise health awareness about things like how to eat well, quit smoking, and more.

But like any other chatbot, SARAH started giving incorrect answers. Leading to a lot of internet trolls and finally, the usual disclaimer: The answers from the chatbot might not be accurate. This tendency to make things up, known as hallucination, is one of the biggest obstacles chatbots face. Why does this happen? And why can’t we fix it?

Let’s explore why large language models hallucinate by looking at how they work. First, making stuff up is exactly what LLMs are designed to do. The chatbot draws responses from the large language model without looking up information in a database or using a search engine.

A large language model contains billions and billions of numbers. It uses these numbers to calculate its responses from scratch, producing new sequences of words on the fly. A large language model is more like a vector than an encyclopedia.

https://medium.com/media/260ffdc0d69d4e1186a4969c41c75c63/href

Large language models generate text by predicting the next word in the sequence. Then the new sequence is fed back into the model, which will guess the next word. This cycle then goes on. Generating almost any kind of text possible. LLMs just love dreaming.

The model captures the statistical likelihood of a word being predicted with certain words. The likelihood is set when a model is trained, where the values in the model are adjusted over and over again until they meet the linguistic patterns of the training data. Once trained, the model calculates the score for each word in the vocabulary, calculating its likelihood to come next.

So basically, all these hyped-up large language models do is hallucinate. But we only notice when it’s wrong. And the problem is that you won’t notice it because these models are so good at what they do. And that makes trusting them hard.

Can we control what these large language models generate? Even though these models are too complicated to be tinkered with, few believe that training them on even more data will reduce the error rate.

You can also ensure performance by breaking responses step-by-step. This method, known as chain-of-thought prompting, can help the model feel confident about the outputs they produce, preventing them from going out of control.

But this does not guarantee 100 percent accuracy. As long as the models are probabilistic, there is a chance that they will produce the wrong output. It is similar to rolling a dice even if you tamper with it to produce a result, there is a small chance it will produce something else.

Another thing is that people believe these models and let their guard down. And these errors go unnoticed. Perhaps, the best fix for hallucinations is to manage the expectations we have of these chatbots and cross-verify the facts.

---

Why does AI hallucinate? was originally published in Becoming Human: Artificial Intelligence Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

Bengaluru-based fintech giant Razorpay is taking a bold step to pave further a better future in India’s digital person-to-merchant (P2M) payments market. It is currently valued at more than $7 billion. It has a strategic plan to go public in the next 2-3 years. It aims to achieve $750 billion in payment volumes by 2030. […]

Waste and pollution are two concerning issues in the cities. Startup Golden Feathers has emerged to address the problem. Radhesh Agrahari is the visionary behind it. His initiative converts chicken butchery waste into wool-like fiber and handmade paper. It is an innovative idea to reducing river pollution. The journey of Radhesh started at the Indian […]

The Paris 2024 Summer Olympic Games is being said to be a landmark event. The reason is not just for the athletes and their quest for medals. There is another reason. It is the use of technological innovations to enhance the experience for everyone involved. Artificial intelligence (AI) and cloud technology are at the forefront. […]