Quantum Computing and Large Language Models

Challenges with Current Large Language Models (LLMs)
• LLMs, like OpenAI, Google, and Microsoft, are crucial in AI, but they consume significant energy for training and usage.
• Larger models like GPT-3 require more computational power, consuming more than an average American household in 120 years.
• Training an LLM with 1.75 billion parameters can emit up to 284 tonnes of carbon dioxide, more than the energy required to run a data centre with 5,000 servers for a year.
• LLMs’ pre-trained nature restricts user control over their functioning, leading to “hallucinations” where the model’s understanding may diverge from reality.
• Current LLMs struggle with syntax, which is the structural arrangement of words and phrases in a sentence.

Quantum Computing and Syntactics and Semantics
• Quantum computing can address these challenges by harnessing the properties of quantum physics like superposition and entanglement for computational needs.
• Quantum natural language processing (QNLP) has emerged as an active field of research with profound implications for language modelling.
• QNLP incurs lower energy costs than conventional LLMs and requires far fewer parameters than classical counterparts, promising to enhance efficiency without compromising performance.
• QNLP uses a better “mapping” between the rules of grammar and quantum physical phenomena like entanglement and superposition, resulting in a deeper, more complete understanding of language.

Time-Series Forecasting
• Quantum generative models can work with time-series data, allowing quantum algorithms to identify patterns more efficiently and solve complex problems related to forecasting.
• A QGen AI model built in Japan successfully worked with both stationary and nonstationary data, demonstrating that fewer parameters were required compared to classical methods.
• By embracing QNLP and QGen-AI, advancements in time-series forecasting can pave the way for sustainable, efficient, and performant AI systems.