Blog

ChatGPT signifies a God-like entity for OpenAI. There is a race among tech companies to attain a lofty concept of artificial general intelligence (AGI) that either matches the cognitive capabilities of humans or surpass it. The top management of tech companies were striving for financial prosperity that could be beneficial to humanity.

Despite the noble intentions, it goes without saying that the efforts of the tech companies put their financial results first and foremost. The altruistic objectives seem to have fallen by wayside. The generative AI boom has become a race to win.

There are various opinions about the timeline to achieve AGI. Anthropic sees it coming by 2027. Altman of OpenAI puts it away ‘by a thousand days’. SoftBank predicts a two-three years’ timeline. All these leaders are not settled about one definition of AGI. Hassabis treats it as software that can perform at ‘human level.’ Altman feels AGI would ‘outperform humans.’ Microsoft calls it ‘nonsensical benchmark hacking.’ AI’s top leaders dance around various definitions of these North Star, while continuing to move towards it. Some draw a gloomy picture by associating it with existential threat. The door is vaguely defined. The consequences are not known.

There is no evidence that any of the existing players have AGI, but they are confident that once there they will tackle the most complex problems facing the world — cancer cure, climate change.

In their attempt to build AGI, the existing players strived towards building the bigger models in the hope that the bigger is better.

The idea should not be to build gods since the benefits are still not certain

As we know, Microsoft has already invested in OpenAI about $13 billion. It also backs BlackRock. Elon Musks XAI is the rival firm for OpenAI. Both Microsoft and BlackRock are investing in XAI to build $30 billion AI infrastructure. Nvidia is the technical advisor to the group.

Microsoft is increasingly developing AI outside OpenAI partnership. It has created in-house AI models that could compete with OpenAI. Since AI development requires huge capital investment in data centers, the US has taken a stargate initiative. The Microsoft-Backed group will be called AI Infrastructure partnership (AIP). BlackRock’s CEO Larry Fink wants to bring pensions and insurers together for such long-term infrastructure projects.

Microsoft BlackRock tied up in 2024 MGX, a UAE investment vehicle that is also involved in Stargate.

They have tied up with energy companies to accelerate the scaling of the project. By

Micro-dramas are vertical dramas on videos under two minutes and are very popular. It will create a $13 billion dollar market globally by 2027. In India, it is becoming mainstream entertainment. Some of the Indian shows produced by KUKU are My Idiot Hero Husband, Secret Billionaire, Study God.

It is bite-sized, emotionally, charged storytelling. These videos are produced in vernacular. Slatforms such as YouTube Shorts, Instagram Reels have noticed this trend.

The trend has picked up after data consumption has become affordable and there are smartphones. The success of bite-sized video content is addictive.

Some players in this market have adopted subscription model. There are subscriptions for different time durations. There are premium subscriptions. In monetizing the content, there could be ads or product placements. In future, brands can be made part of the narrative.

The production cost of micro-dramas is very low. A single micro-drama takes Rs.5000-Rs.10000 per episode to produce. There is real cost advantage in the distribution model itself — as these are sent direct to platform, the middlemen are eliminated. They thus reap quicker returns on investment. AI-driven content can further bring down costs.

Masters’ Union School of Business has been founded in 2020. It is located in Cyber City area of Gurugram. Its inaugural programme was the 16-month Post Graduate Programme in Tech and Business Management. In addition, it runs three more programmes — PGP in Family Business Legacy Management, UG in Tech and Business Management and UG in Psychology and Marketing.

The students at this school get a salary package equal to the IIMs or better than that. The faculty consists of CXOs. The students run investment fund. The Course also teaches students how to code. CXOs as teachers make the students hands-on.

Master’s Union has started the School of Emerging Technologies and has started courses in Data Science and AI. It is tied up with Illinois Institute of Technology. It is a 3+1 dual degree with 3 years in India and the 4th year at Illinois Tech. For its India track, the institute has tied up with IIT at Madras and Guwahati. The whole course costs Rs. 1.03 crore.

The admission is on the basis Jee (Mains) scores and acadamics and extra-curricular activities.

Masters’ Union has started another business school — Tetr College of Business. Its course has 8 semesters. It is taught at different locations. It starts in Dubai, followed by India, Singapore, Malaysia, Ghana, US, Argentina and Europe. The last 8th term is that of internship.

Bitcoin reserve was established by an executive order of the US President on March 6, 2025, to avail of the first-mover advantage.

The order describes it as digital gold. Since there is a finite supply of BTC, it gives a strategic advantage to be the first among nations to create its reserve.

The US Departments of government would transfer all BTC they hold and have forfeited from criminal and civil proceedings treating them as US Digital Assets Stockpile (DAS). The additional BTC will be acquired in a way that does not cost taxpayers.

The US government by creating a reserve, treats BTC as a viable asset. By calling it digital gold, it can be used as a means to diversify assets holdings which can hedge against inflation and be the store of value (like gold reserves). BTC is treated as first among equals in all cryptos. The US holds 2 lac BTCs (17 billion) mostly obtained through criminal forfeitures.

The US actions inspire other nations to institutionalize its use. It could lead to a big shift in the global financial system. It could lead to infrastructure such as exchanges and indices. It could enhance the value of BTC.

David Sacks, the Crypto Czar, point out that BTC has $ 2 trillion market cap. It is the most secure — it has not been hacked so far. The executive order is to be converted into law. BTC is not just a technological opportunity. It will provide financial leadership to the US in 21st century. The US could reduce its deficit (without raising taxes) if the BTC appreciates in value. It may strengthen US dollar.

Cryptos are not yet regulated and are extremely volatile. SBR attempt tries to legitimize BTC. The President and his close associates may hold significant quantities of crypto portfolio personally.

India has brought digital assets under the prevention of Money Laundering Act, 2023. The service providers have to register with the Financial Intelligence Unit (FIU). Taxation policies discourage holding cryptos. India’s digital financial system consists of UPI, Aadhar-enabled payments and the digital currency. The RBI feels cryptos can undermine monetary policy, create fiscal risks and encourage capital flows.

Coinbase, the US-based crypto exchange, has been given green signal to re-enter crypto trading by the FIU in March 2025. It is to be seen how long India keeps cryptos at an arm’s length.

2024 is considered to be the second-best year for Indian cinema in terms of box-office collections — the industry earned Rs.11833 crore, slightly short of 2023’s Rs.12,226 crore (Ormax Media Report).

Despite this, cinema footfalls do remain below the pre-pandemic levels. In 2024, the drop in footfalls was 6 per cent. The decline could be attributed to digital platforms where movies get released immediately after the theatrical release or in some cases at the same time. In addition, OTT platforms are ad free, and the content can be consumed anytime.

Cinema halls do excessive advertising — in a recent case a multiplex screened a movie 25 minutes after its scheduled time, and the audience sat through the ordeal. The Consumer District Commission received a complaint and penalized the theatre.

The HC stayed the ruling. The tickets must specify when the actual movie will begin.

The issue is how to maximize the revenue without irritating the audience. The duration of advertising time could be limited, say screening time of 10 minutes if the audiences are to be kept happy.

Brands should focus on creating concise and visually compelling ads that resonate emotionally.

Indian cinema advertising grew by 10 per cent in 2024. It is expected to grow by 9 per cent in 2025. The ad revenue earned is Rs.950 crore. It is 8-10 per cent of cinema revenues.

Advertisers get the benefit of a large screen impact in cinema advertising.

With big blockbuster’s, the theatres tend to screen ads for 30-45 minutes at each screening and weekend shows.

This is a risk that they cannot afford to take .

Excessive advertising does infringe upon consumer rights. Theatres should tap other avenues such as digital displays, interactive kiosks and in-lobby promotions.

Big Tech (Google, Facebook, X, Amazon, Microsoft) not only connect us with others, but also shape how we think, consume and interact with the world.

Capitalism always favours laissez faire, since deregulation is assured to be dynamism and freedom. On the contrary, unrestrained power does not ensure freedom.

Big Tech acts as gatekeepers of information, commerce and public discourse. They have the potential to undermine competition and free enterprise. However, there are voices that regulation could stifle innovation.

We have adopted a digital lifestyle which generates voluminous data, and it is controlled by a few players.

Startups are vulnerable to the digital eco-system. This vulnerability could be weaponised.

The availability of data with Big Tech poses a challenge to state’s exclusivity and sovereignty.

Canada and Australia try to control digital news distribution. Legislation have been introduced to compensate for the content. To retaliate, Facebook blocked government ads in Canada. In Australia, Big Tech retaliated by banning the sharing of news and links.

Big Tech extends US state power globally. The US gets extraterritorial reach. Much of the world’s communication flows through the US. SWIFT’s allows US agencies access to its database. PRISM (NSA) taps user data. Huawei has been placed on US entity list.

India has to strengthen data protection (implementing Digital Personal Data Protection Act, 2023). Some critical data should be stored within the country. Innovation must flourish, but not at the cost of undermining a country’s interests.

Ilaya Sutskever, co-founder of OpenAI and now head of Safe Superintelligence in a recent conference in Canada expressed his opinion about the current AI systems based on pretraining. According to him, pretraining, as we know it, is going to end unquestionably. LLMs learn patterns from vast amounts of unlabelled data sourced from internet, books and other sources. Already, the data drawn has reached its peak, and there is no more data to be drawn. He compared this situation to fossil fuels — oil is a finite resource. Similarly, internet contains a finite amount of human-generated content.

Existing data can still take AI development farther. However, the industry is tapping out on new data to train on. There will be a shift away from the way the models are trained.

The more a system reasons, the more unpredictable it becomes. AI systems which play chess are unpredictable to the best human chess players.

Sutskever is optimistic about agentic models, which can understand things from limited data. Apart from being agentic, future systems will be able to reason. Today’s AI is mostly pattern matching based on what the model has seen. Future AI systems will be able to work out things step by step, more like the way we think.

Sutskever feels that futue models would depend not only on text but also on muli-modal data , including computer vision

The world is after attaining AI supremacy. Already a low-cost DeepSeek –R1 has caused a global sensation. India too proposes to develop indigenous AI models. India intends to support compute power by a stockpile of 18,693 GPUs. There is a proposal of 40 per cent subsidy to developers. It will reduce per hour computing costs.

The state-of-the art global models too suffer from latency and slow response time. They are less efficient than SLMs. LLMs are good performers. Distilled models (DMs) stand in-between. They are relatively less efficient than SLMs.

India cannot remain confined to one type of model. India needs foundation models and LLMs for advanced research in defence, national security and atmospheric studies- – to predict adverse national phenomena, DMS may be used. To answer a query of a farmer in his native language, a model of small and medium size having NLP capabilities could be used.

Generative AI models are capital intensive. Using GPT-4 for finding chemists selling surgical masks is a waste of resources.

LLMs such as Llama-2 and DeepSeek follow ‘open weight'(OW) system of disclosure. It permits users to fine-tune the parent model for customized requirements. OW also enables researchers to test fairness and safety features of a model.

OS models such as Mistral and Falcon not only disclose weights and codes but also information on datasets. Users can do unlimited modification to the parent model and create new models. India’s A14 Bharat model is a pure OS model.

Open Weight models do not disclose training data. They can face lawsuits in markets.

Since its release in November 2022, ChatGPT has stirred the users so much that they wonder about the capabilities of Large Language models (LLMs) in particular and AI in general. It is difficult to come across someone who has not experienced the power of ChatGPT. While all these tools such as GPT, Gemini or Claude are powerful with hundreds of billions of parameters and pre-trained on vast corpora of text, they are not omnipotent. These models fall short for specific tasks — however these models can be used for speafic tasks by fine-tuning them by using techniques such quantizatron and LORA. There are some libraries for fine tuning.

Fine-tuning is an expensive process, especially a model with a large number of parameters. Models with less than 10 billion parameters can be fine tuned without any significant infrasructure changes. For larger models, we require approximately 1.5 terrabytes of GPU vRAM. It is equivalent to a cluster of 20 Nvidia A 100s, each with 80 GB of vRAM. This set up costs $ 4 lac. The assumption is the hardware is available.

Alternatively, one can use one of cloud providers (AWS, Azure or GCP). This approach too is expensive. An hour of using A 100 GPU on AWS costs $40. If a model is fine-tuned on 20 GPU for 5 days, it would cost about $1 lac.

That is why researchers use smaller LLMs with less than 10 billion parameters. A Mistral can be fine-tuned using Nvidia A10 on AWS. It takes 10 hours, costing less than $20. However, the model requires quantization.

Quantization converts model’s parameters to low-precision data types — 8-bit or 4-bit. This reduces memory consumption and speeds up execution. All 32-bit values are mapped to a smaller range of finite values — 256 for 8-bit conversion.

Another technique LoRA is low-rank adaptation. Here model’s weights are updated using matrix dimensionality reduction. Transformers, as we know, rely on matrices. Here parameters are adjusted within these matrices. In LoRa, two smaller matrices are created for updation.

A Matrix with 1000×1000 parameters (totaling 1 million parameters) can be decomposed to 1000×1000 multiplied by 100×1000 matrices. This reduces parameter count to 2*100k (a reduction of 80 per cent in parameters). The approximation is less precise but still it improves memory and computational efficiency significantly.

Quantization and LoRA can be used in combination. It is called QLoRA.

To begin fine-tuning anew, Unsloth Python library is used. After pre-training an LLM, there is supervised fine-tuning (SFT). The tools used are SFT and PEFT (Parameter Efficient Fine-tuning) from Hugging Face. In addition, LoRA and quantization can be easily applied by using BitsAndBytes (by Tim Dettmers).

Lastly, after pre-training and supervised fine-tuning, the model is informed which generated outputs are desirable and which are not. It is called preference optimization. The techniques used are RLHF — reinforcement learning from human feedback and Direct Preference Optimization (DPO).

In March 2024, a new technique has emerged called Odds Ratio Preference Optimization (ORPO) combining supervised alignment.