Censorship is Killing ChatGPT
AIM Events
Hosting the World’s Most Impactful AI Conferences & Events. For Brand collaborations write to [email protected]
What happens when you try to integrate social values into chatbots? They break down. The RLHF (Reinforcement Learning Human Feedback) used to create a chatbot aligned with human values has stymied the true potential of the platform. The phenomenon is called the alignment tax of AI models.?
A paper titled ‘Scaling Laws for Reward Model Overoptimisation’ delves into the phenomenon where RLHF preferences lead to biased models that impede the true performance of the models. Hence, every time an LLM is fine-tuned, it hinders its overall functionality and loses its performance.?
This month, a new development challenged the position of ChatGPT and was antithetical to the idea of RLHF. A creator named Eric Hartford released an LLM model called WizardLM-7B-Uncensored on Hugging Face. The model gained prominence for its improved intelligence and creativity as it removed censorship from its training data.?
Uncensored models, or models that lack alignment seemed to perform better than aligned models like GPT-4, PaLM and others.
Besides, WizardLM-7B-Uncensored has shown the need for uncensored models for scientific exploration, freedom of expression, composability, storytelling, and even humour. The idea of uncensored LLMs sounds amazing, but it is not error-free. The debate of exercising freedom of speech and expression and not offending someone is as much part of chatbots as it is in our society.?
The doomsayers always raise objections to the use of uncensored chatbots for malicious purposes and holler about the danger of it falling into the hands of earth destroyers. Wow! This approach is too extreme; and thus, the debate between aligned and uncensored LLMs is going to continue.
Read the full story?here.
Jensen Huang: The Steve Jobs of Chip World
NVIDIA has become the 7th company in the world to reach a trillion dollar ($1 trillion) market. This is a big achievement for the 30-year-old company. However, to reach this stage, in its three-decade journey, the company has taken many bold steps which went against the market trend.
In a recent speech to the graduating students of National Taiwan University (NTU), Huang shared many instances when the company rode against the tide and bet on counterintuitive ideas. For instance, in 2010, when every company was fighting to be a part of the phone market, NVIDIA, instead of trying to fight for a market share in the segment, retreated from it.?
领英推荐
Read the full story?here.
Life Transforming Mantras from Tech Leaders
NVIDIA's founder and COO, Jensen Huang, delivered a memorable commencement speech at NTU, encouraging graduates to seize opportunities fearlessly and learn from failures. Throughout history, various tech pioneers have delivered impactful speeches, shaping lives by imparting invaluable wisdom. We have compiled some interesting quotes from a few tech greats here. Take a look:
Read the full story?here.
Microsoft Minus OpenAI = ?
In 2019, software giant Microsoft played a huge bet on OpenAI with its first investment in the company. Obviously, the tech giant was betting big on an idea where it saw success for itself. Since then, the company has invested about $13 billion in the startup.?
Interestingly now, Microsoft, which hasn’t launched any product in many years, has suddenly become active in the AI segment. It has integrated Codex and GPT in every single offering announced recently. And the company has also enabled generative AI capabilities in Windows and Office 365. So, while other tech giants have in-house AI products, Microsoft’s AI empire is being built on OpenAI.
Read the full story?here