The AI Success Mantra

There’s no one surefire way to succeed in AI. It can be achieved by employing various strategies and methods, and what works for one, may not work for another, thus making each journey unique. The same holds true for chipmakers and hardware companies like NVIDIA, Qualcomm, Huawei, Intel, AMD, Cerebras, and others. In this AI hardware race, there is no single winner — each of them follows a unique approach.

While NVIDIA and Huawei seem to be pouring their heart and soul into making AI supercomputers, Qualcomm is busy experimenting with bringing AI edge use cases, which is similar to Apple’s approach of developing and promoting on-device AI models.

And, while Apple hasn’t officially announced that it is building GPT-like models for its devices, Qualcomm has said that a hybrid-AI approach is essential for reducing the cost of running AI models on a single device.

Interestingly, Intel has also decided to strides into the GPU market, getting closer and closer to NVIDIA. Pat Gelsinger, the CEO of Intel, believes that the company now has a simpler roadmap for its GPU and accelerators, and plans to onboard Falcon Shores 2 , the second version of its AI superchip, by 2026.?

This seemed like an almost good approach by Intel until NVIDIA decided to jump on the bandwagon. At NVIDIA’s recent financial call, CEO Jensen Huang said that the company plans to introduce L40S , a GPU specifically designed to fine-tune and inference. Given that people are already using NVIDIA H100s for training, shifting to Intel’s Xeon processors may be a big leap to take.?

This move by Intel left the industry astounded as it was the first time the company was taking a GPU-only approach. Intel’s Gaudi2, which worked 2.4 times faster than A100 , and almost matched the capabilities of the H100 Hopper GPU, seemed to question NVIDIA’s dominance and strength. The third competitor, AMD, will also be set with its Instinct MI300 hybrid CPU-GPU accelerators by the end of 2023.

Intel believes that there are two AI markets — one that deals with infrastructure for training AI models (for which the company has Habana Labs Gaudi2 ), and the other for fine-tuning and inference (for which it trusts its Xeon Max 9480 processor to be on par with NVIDIA GPUs).

More leaks about the upcoming 14th Gen Meteor Lake processor also suggest that Intel’s CPU might have a DDR5 memory, which is also similar to Apple’s M2 chip design. Moreover, it is also expected that AI will play a major role in the Meteor Lake CPUs.?

Read: Intel Soon to be on Par with NVIDIA

Similarly, Cerebras Systems, the company that boasts one of the largest supercomputers in the world, said that it is the first company that is able to train AI systems that do not rely on GPUs . The Silicon Valley-based firm has trained all its models using 16 CS-2 systems on its Andromeda AI supercomputer.

Everyone's a winner in AI

Meanwhile, in China, Huawei is hell-bent on headlining the country’s AI mission . Liu Qingfeng, the founder and chairman of iFlytek, a company planning to release ChatGPT competitors in October, claimed that Huawei’s GPU capabilities are now on par with NVIDIA’s A100 GPUs .?

Though Huawei hasn’t yet officially announced that it is building GPUs for AI development, iFlytek’s comments confirm that Huawei’s Ascend 910 AI accelerators are indeed being used by AI companies in China. It is clear that the USA's effort of putting sanctions on selling chips to China has turned into a self-reliance mission for the country.

Every company is planning to carve its niche in the AI domain, and is thus supporting the AI revolution with its unique strategies. Some go for GPUs, some CPUs, and some focus specifically on edge use cases. AI has helped companies realise that in the end, it is not just a competition — everyone should be a winner in AI.

AIM RESEARCH >>

Analytics and Data Science Jobs in India 2023 by AIM Research & Great Learning

• The much-anticipated report on Analytics and Data Science Jobs in India 2023 is out now.?Compared to last year, we observed a dip in the number of open roles for data science experts in India.
• City-wise, while Pune, Hyderabad, and Delhi NCR are burgeoning with opportunities for data scientists, Bangalore and Mumbai have seen a tapering demand.
• Sector-wise, BFSI, healthcare, and retail & CPG are leading the charge, with MNC IT and KPO sectors presenting a plethora of job openings.
• BFSI Sector Dominance: In 2023, the BFSI sector leads the pack, accounting for a whopping 36% of data and analytics job openings.
• Salary Insights: A significant chunk of job positions in 2023 falls in the 6-10 LPA bracket for analytics and data science specialists.
• Experience Matters: 34% of job openings are tailored for professionals with 2-5 years of experience. While mid-level roles are on the rise, senior-level positions have seen a dip.
• IT/BPO/KPO Sector: Contrary to last year, 2023 sees an 18pp decrease in the availability of analytics and data science jobs in this sector.

Access the Complete report here>>

TOP STORIES OF THE WEEK >>

How Many GPUs is India Buying from NVIDIA?

NVIDIA’s partnership with Reliance and Tata , alongside its vision to upskill Indian IT employees, is just the beginning of the Indian AI revolution. CEO Jensen Huang said Reliance will be using NVIDIA’s infrastructure to create customer-centric AI applications and services for its 450 million Jio customers and provide energy-efficient AI infrastructure to scientists, developers and startups across India.?

Five years ago, Huang hand-delivered the first NVIDIA DGX AI supercomputer to a startup, which was none other than OpenAI . If it took them about five years to reach where they are, how much time will it take for Reliance or Tata to reach OpenAI-level of success in AI??

Read to find out .

Hallucinations are Bothersome, but Not All That Bad

AI spewing made-up facts is inevitable since it is not a search engine or database. But, at the end of the day, it’s still a technological revelation. Meanwhile, despite its prevalence in the media, tech blogs and research papers, the word ‘hallucination’ is inappropriate for AI, many argue. OpenAI CEO Sam Altman recently said that hallucinations in generative AI models can be, “a feature, not a bug”.

In its latest edition of the ‘Schizophrenia Bulletin’, Oxford researchers published a piece titled, ‘False Responses From Artificial Intelligence Models Are Not Hallucinations ’. Guess what, they are not the first ones to find the term ‘hallucination’ inappropriate while referring to a piece of technology.

Read more here .

PEOPLE & TECH >>

Bringing ‘Common Sense’ to Machines with Former Google DeepMind Scientist

Hailing from Pune, former Google DeepMind research scientist Tejas Kulkarni co-founded Common Sense Machines , a Boston-based AI startup aiming to revolutionise 3D generation AI platforms. Kulkarni believes that the missing piece to get to AGI is the part where machines have the ability to think, or rather, have ‘common sense’.

Speaking to AIM, Kulkarni said, “We were the first to build an image to 3D model at our scale — nobody had created it. There isn’t a GPT of this world yet.” He further spoke about AI art generators like Midjourney, his journey into the AI world, and the current gaps in the transformer models that prevent machines from thinking on their own.

Read the interview here .

AIM Shots?>>

• Analytics India Magazine (AIM), your favourite platform for a wide-angle world view on analytics, data science, and artificial intelligence, has officially announced its foray into the US .
• OpenAI now has a brand-new office in Dublin. The objective is to collaborate with the Irish government and support their national AI strategy and work with industries and startups.
• AWS India has entered into a strategic MoU with ISRO and IN-SPACe to advance India’s space capabilities through cloud technologies.
• Beverage giant Coca-Cola is using AI to develop a new flavour: Y3000 Zero Sugar . This unique flavour aims to offer a “taste of the future ”.
• NVIDIA has announced that it is set to release TensorRT-LLM , an open source software that promises to accelerate and optimise LLM inference, ? in the coming weeks.
• Microsoft researchers have open sourced phi-1.5 , a cutting-edge language model of 1.3 billion parameters that outperforms Llama 2’s 7-billion parameters model on several benchmarks.
• After a six-month beta, Adobe , the design software giant, has leapt towards customer-centricity, launching a web application for Firefly, a suite of generative AI models.
• Redis, Inc has announced the integration of Redis Enterprise Cloud ’s vector database capabilities with Amazon Bedrock , a service designed to facilitate the creation of generative AI models.