ChatGPT, Bard, Bing, Claude: How 2023 Marked the Age of Generative AI

The AI wars technically commenced on 30 November 2022, but have had a profound effect on every part of the tech industry in 2023. OpenAI, a seemingly inconspicuous AI non-profit, had barely attracted media attention since its inception in December 2015, despite being founded by AI luminaries and Silicon Valley notables such as Sam Altman and Elon Musk.

A substantial $1 billion investment from Microsoft also escaped widespread notice as OpenAI transitioned from its non-profit roots to a unique capped for-profit model, a move that played a part in the upheaval it experienced in November 2023.

The Tau of ChatGPT and Microsoft

The release of GPT 3.5 at the end of 2022 triggered a state of high alert at Google, which had been viewed as the foremost authority in AI for two decades. Interestingly, Google was the birthplace of transformer models and LLMs, technologies that have shaken the industry. ChatGPT, based on GPT 3.5, demonstrated that Google’s dominance, backed by its vast financial and talent resources, was not only challengeable but could be surpassed. The prospect of Artificial Generative Intelligence (AGI), once thought to be decades away, suddenly seemed achievable within the 2020s. ChatGPT soon became the fastest product to achieve a million active users.

In January, Microsoft introduced Bing Chat, which evolved into Microsoft Copilot within the year. Utilising the framework of OpenAI's then-undisclosed GPT-4 thanks to Microsoft’s investments into the startup, Bing Chat represented a significant leap in search technology, rivalling the impact of Google’s original PageRank algorithm. Satya Nadella, CEO of Microsoft and a key figure in the company's cloud computing and Bing Search development efforts, asserted his intention to challenge Google's dominance, leveraging

OpenAI's generative AI capabilities.

Nadella's focus isn’t merely on market share—a modest single-digit percentage would suffice, translating into billions in value. His real aim is to capture mindshare, showcasing the strengths of a Microsoft that has risen to become the world’s second-most valuable company, trailing only Apple. Microsoft Azure, the infrastructure powering OpenAI’s transformer models for GPT 3.5 and later GPT-4, stands at the core of this strategy.

Nvidia’s Rise as the Chip Provider of the Gen AI Age

Meanwhile, Nvidia, the chip giant, experienced rapid growth since the onset of the pandemic. As of this article’s writing, it has become the most valuable semiconductor player, thanks to its H100 and A100 series AI compute engines, crucial to Microsoft Azure's operations. Nvidia's latest offering, the Grace Hopper supercomputer H200, priced at around $300,000 per unit, will have Microsoft as a primary customer. However, Google's Cloud and AWS are also vying for these chips, essential for training foundational models for generative AI startups aiming to rival OpenAI.

Demand for these chips outstrips supply, prompting other semiconductor companies such as AMD and Intel to accelerate their efforts in chip development to support the burgeoning AI sector. Microsoft Azure, AWS, and Google Cloud are developing their own ASICs to become more competitive. Microsoft’s Maia, Amazon’s Tranium 4 and Google’s new TPU v5e and A3 GPUs enable these cloud giants to run specific workloads which is beneficial for their parents and also their clients.

Generative AI's influence extends to on-device processing. Qualcomm's Snapdragon Elite series promises to bring this technology to devices next year. Intel's Meteor Lake architecture, representing a significant shift in x86 CPUs, includes a robust NPU for on-device generative AI processing, a path also followed by Apple with its Neural Engine buzzword and AMD with part of its mobile lineup. Nvidia and AMD are also reportedly designing ARM-based processors for PCs which likely will have a huge generative AI spin in the near future. Qualcomm’s Snapdragon 8 Gen 3 went big on on-device processing along with MediaTek’s Dimensity 9300, which will debut in the Vivo X100 Pro. Google, too, is advancing in this direction with its Tensor G3 processor and Gemini Nano model for on-device foundational AI.

Google and Other ChatGPT Challengers

Google, initially caught off guard, has made significant strides with its Bard AI, built upon its Palm 2 LLM model. Despite initial challenges, Bard has been steadily integrated into Google’s suite of services, including YouTube, Search, Google Workspace, and Gmail.

Furthermore, Google’s Gemini project is set to advance these efforts with the upcoming Gemini Ultra claimed to surpass even GPT 4. When GPT 3.5 came out there were reports that Sundar Pichai, the Alphabet CEO, had asked Sergey Brin to come back to help with the project. It is said that the Google co-founder actively wrote code for Gemini and worked long hours with the Google DeepMind teams.

Other significant players include Anthropic, with its chatbot Claude, and Inflection AI, whose chatbot Pie is gaining recognition. Anthropic has raised billions from Amazon and Google and was founded by an OpenAI alumnus who had worked on GPT 3. Inflexion AI was founded by Mustafa Suleyman, the Google-owned DeepMind co-founder and Reid Hoffman, the LinkedIn co-founder who was till recently on the OpenAI board. There is some serious pedigree here.

These companies’ efforts, along with rapid advancements in image generation technologies such as OpenAI’s Dall-E, Stable Diffusion, and Midjourney, are transforming the landscape of content creation. Adobe has embraced generative AI with its Firefly service and a key partnership with Google is enabling it to infuse Gen AI into iconic creator tools such as Photoshop. Apps including Canva are also integrating these technologies, and generally the way we create videos or graphic art is changing dramatically — the process is becoming easier, but it is also becoming harder to tell what was created by a human and what was done by AI.

In India, the AI scene is also evolving rapidly. Ola has announced Krutrim, its foundational model, which reportedly outperforms GPT 4 in Indic languages and rivals Meta’s Llam 2. The company is also venturing into semiconductor development and datacentre architecture, with plans for a uniquely Indian AI ecosystem by 2025.

The generative AI landscape is undoubtedly dynamic and rapidly evolving, with 2023 marking a pivotal year in its advancement and adoption across various sectors and regions.

AGI Paranoia Taking Over?

2023 was also the year in which the world took note of the dangers of AI. Geoffory Hinton, the legendary AI researcher who had worked with OpenAI’s Ilya Sutskever to create AlexNet, left Google to able to talk about generative AI. Then of course, the coup at OpenAI in which four board members, including Sutskever, voted to fire Sam Altman as its CEO in November stemmed from a fear of AGI (Artificial General Intelligence) emerging sooner than we’re prepared for it, and having the potential to drastically harm humanity if development wasn’t slowed down. OpenAI staff revolted, and with the backing of major investors including Microsoft and Khosla Ventures, Altman is back in charge with a new board, and Sutskever’s future at the company is clouded.

Google’s lethargic approach towards commercialising LLMs has also reportedly been due to its fear of this technology -- but its hand was forced by OpenAI. This technology is also a threat to its core search business. Elon Musk’s X has its own LLM called Grok which also leverages his assets at Tesla and SpaceX, and has also reportedly been developed because of his fears of AGI wiping out humanity. Regulators in the US and EU are looking at the space very actively, and now the EU has also passed an act on accountability and regulations for AI -- though it is still being worked upon, as many view it as an innovation killer.

New Gen AI-Infused Gadgets

Apple, notably, is one company that hasn’t had much to show in this space -- but 2024 could be the year that even Apple dives in. 2023 has been the beginning of a new era of technology, and we’ve already seen it applied in new hardware like the HumaneAI pin which leverages LLM models to respond to queries and do things. Sam Altman was in talks with former Apple Chief Design Officer Sir Jonathan Ive to create new AI-first hardware -- that could be something we see in 2024 even as a concept. Meta has already updated its Ray-Ban glasses with multimodal capabilities and one can be sure that Apple will reveal some AI-related surprises as its Vision Pro headset ecosystem matures. A new age is dawning, and AI is already changing the way we write, read, converse, and create.