Nvidia, the chip supplier, mainly to the AI sector confirmed in September 2024 that its new Blackwell chip had faced some production problems, affecting Nvidia’s shares which fell by 8.4 per cent. It happens to a company that has shown spectacular growth in past. And its growth forecast too beats expectations. This justifies the adage in Silicon Valley that hardware is hard.

The Blackwell delay is a temporary hitch. The overall margins are enormous. The surging demand for its chips will continue for many more quarters. That Blackwell is a little late is not an issue. The existing products sell like hot potatoes. The production issue is sorted. It was an issue related to operations rather than the chip design. It was not an issue where the product will be back-to-the-drawing-board type. The company has streamlined its operations. It pledges to update its flagship hardware at least once a year. They hope to have a great next year.

Investors should better be concerned about the biggest buyers of Nvidia chips making their own components. They contribute 45 per cent to Nvidia’s revenues.

The issues may not become obvious in a couple years. Even if AI is overhyped, it will be detected after many more billions being spent on Nvidia chips. Till then, there is no cause for investors to overreact.

Of course, hardware is hard. Keeping Wall Street happy is even harder.

AI is transforming our world. ChatGPT and Gemini are known for generating human-like text.

Researchers have come across AI models trained on data that is generated by previous versions of the models. It is moving away from the original data distribution. The output could be distorted and unreliable. It is called model collapse.

As we know, AI models are trained vast volumes of data. It has data scraped form internet. The data is, to begin with, generated by humans. The model learns patterns from this data. The future generations of the model use data from two sources — human-generated data and data generated by the previous models. There is degradation in the output as the data quality successively goes down. Each version dilutes the original detail. The final output is a hazy and less accurate description of the world around. It is a slow process, but it happens. It is like making a copy of a copy, and so on. There is a loss in output. It is inevitable.

The content is less creative, more stereotyped and less useful.

It is not a limited problem. It has far-reaching impacts. There is decline in the efficiency of the model. The model becomes less reliable. It may commit costly errors. The issue of bias raises its ugly head.

One solution is to restrict the model’s training on human-generated data. However, of late, much of the internet content is model generated. And there is an issue of distinguishing human-generated content and AI-generated content.

At times, AI-generated content mimics human-generated content.

The use of human-generated data has its own problems — there are ethical and legal issues. The first-mover advantage is enjoyed by the pioneering models, as there is less contaminated data for training. Thus, early adopters are at an advantageous position.

It is crucial to have access to human-generated data. This has to be balanced with the rights of those whose data is being used. There should be cooperation at the industry-level. Models should be continuously exposed to fresh human-generated data.

Though AI models are powerful, they are still dependent on the quality of data they are trained on.

Though model collapse poses a problem, we can overcome it by implementing the right strategies.

Telegram founder and CEO Pavel Durov was detained at an airport near Paris. It has repercussions all over. Some called it an assault on free speech and innovation.

There are nuances here. Telegram has been ultra-lax about the content oversight. There are allegations of child sexual abuse (CSAM) material on the platform. Other social media platforms spend considerable time and resources to ban such material. There are alleged drug trafficking and money laundering issues.

Telegram is a powerful platform with 900 million monthly users. There is no overseeing worth the name. It has minimum intervention policy (resulting into low operational costs). Undesirable groups may be using the platform not for its secrecy but for its ‘anything goes’ approach.

Telegram calls it absurd that the owner of the platform is held responsible for the abuse of the platform. It says it abides by the EU laws, including the Digital Services Act. Many others do not see any absurdity if someone is held accountable for criminal activity. Telegram is not end-to-end encrypted. Most chats on the app use client-server encryption. If it chooses, Telegram can access message contents. Much of the content is on public channels. Telegram’s ‘secret chat’ feature is end-to-end encrypted. It is not active by default. It is not used for regular communication. Telegram does not offer real privacy, but only creates an illusion of it. It has technical means to monitor content. It chooses not to use this capability.

Telegram has allowed a free rein to all kinds of activities on its platform and is not immune to its consequences. Digital world is too subject to regulation, just as the physical world is. There cannot be a regulatory vacuum for social media platforms. Europe and Britain are on the right track by either enacting or proposing to enact regulatory laws.

Telegram CEO Druv has been charged with multiple criminal offences. Though it is unusual to target executives for crimes committed on their platforms, Paris prosecutor noted the platform’s ‘near complete’ lack of response to legal requests for cooperation. The 39-year-old CEO posted bail of 5 million Euros after being questioned by a judge on 28th August 2024 and following four days of police custody.

The founder-CEO has acquired multiple citizenships. Druv has dual citizenship of Russia as well as France. His arrest has upset Russia. France treats his as a part of investigation, and not as a political move. When Telegram was run as a free speech absolutist, these passports protected him. Once he wrote in Instagram, ‘To be really free, you could be ready to risk everything.’ That risk appears to have caught him up (despite the passports from Russia, France, UAE, Saint Kitts and Nevis.)

The French charges are against the platform, and not against Durov personally. It serves as a warning to other Big Tech companies. So far social media enjoyed protection across jurisdictions called safe hurbour. Its basic premise is that platforms cannot control the posts on their site. They should not be held legally liable for any objectionable content they host. The only condition is that they should be ready to take such content down when flagged by the government or various courts. Safe harbour is viewed as a basic text of allowing freedom of expression. In the US, this special protection is available to social media under Section 230 of the US Communication Decency Act on par with Section 79 of India’s IT Act, 2000 which classifies social media as intermediaries, and shields them from legal action.

In India, certain officials of social media can be sued if their platforms violate the norms. Under the Rules of the IT Act, 2021, social media companies with more than 5 million Indian users have to appoint a chief compliance officer who could be held criminally liable if the platforms do not adhere to the government’s take down requests or violate other norms. However, the government has not exercised this power so far.

The IT Act will be followed by the Digital India Bill. The government is expected to consider whether safe harbours should be available to the social media.

There is data flood on internet. It becomes difficult for organizations to sift the data that is useful to transform their operations from a huge pool of data. Generative AI has made the data available on internet even bigger, though much of the data is useless for any organization trying to address a specific problem.

Of course, there is a need for big data for business analytics and to get AI solutions. Still, the data inundation in this digital age is fearsome. It becomes difficult to keep the company afloat in this data deluge. Larger storage is made available to deal with the deluge. There is lot of cold data not immediately required for data modelling. It may be used in future to build more efficient AI learning algorithms.

Internet produces several zettabytes of data, and the capacity to store these zettabytes falls short. Data increases at a CAGR of 25 per cent, causing a larger deficit in storage space. In the coming three years, the data could rise to 175 zettabytes. One zettabyte is equal to one billion terabytes. A terabyte is 1000 gigabytes (GB).

How to make sense of this data? Organizations use analytics. The team has experts in statistics, data science and number crunchers. They search for a pin in a haystack. Maybe, they deal with junk, and more junk is likely to be added as time rolls on. Data cleaning has become a humongous task. Much data remains floating on servers without being put to any use. It is not that the data is dirty. The issue is that the data is old. And data ages quickly.

Data storage requires mega data centers. This approach is not sustainable. There are high costs, and there is power consumption.

It is necessary to delete data that has not been used for several years. Let useless photos, files, conversations go offline forever. It is data purging. It requires corporate courage to purge data that is too old to be used. It limits useless computing and storage. The focus is on using relevant and real-time data. The analytical conclusions reached on the basis of such data are actionable immediately.

Indian fashion is being reinvented. There is democratization of design. In past, designs were for the elitist but are now wearable and simple silhouettes.

Indian fashion designers from the 1980s such as Tarun Tahilani (now 62) and JJ Valaya (now 56) have adapted Western couture to Indian sensibility and craftmanship. Tahilani is a Wharton School business graduate and Valaya, a CA. Tahilani set up Ensemble in 1987 in Mumbai and Valaya an eponymous store in 1992 at Sultanpur, Delhi. Much water has flown through the Ganga and the Jamuna since then, and we are on the cusp of another change.

Both these designers gathered like-minded friends — Rohit Khosla, Abu Jani and Sandip Khosla (who costume designed for Umrao Jaan, 1981). There was Anuradha Mafatlal, Anita Shivdasani, Sunita Kapoor and Neil Bieff. They chose a colonial building at Lion’s Gate, Mumbai to showcase about 80 outfits in December 1987. They utilized the services of weavers and embroiders of Mohammed Ali Road. Anarkalis were inspired by Mughal miniatures. French designers thronged there to buy the clothes. These events were closed-door, with no invites sent to buyers.

Valaya was a pioneering batch student of NIFT, New Delhi. He is the founding member of the Fashion Desigh Council of India (FDCI). It was set up in 1998. He is the first to break ice at Paris Haute Couture Week, 2001. He had a solo show at Delhi in early 1990s. He took it to Dubai, Hong Kong, Singapore and London. By 1996, they had a luxury store that went beyond women’s wear and men’s ware. It showcased home decor, furnishings and artefacts. There was an art gallery too and a fine-dining restaurant.

India Couture Week, 2024. Tahilani broke the shackles of elitism by doing a repeat show for those who missed it in the overcrowded hall.

Valaya feels there is no room for resting in the field of fashion. He reset his vision to contemporary trends.

Fashion cannot serve only the rich and elite. It must find a younger voice. Tahilani now experiments with dhoti-saris and button-downed collared cotton shirts. Fashion labels can sustain only if they find a wider market.

There was liberalization in India in the 1990s. It gave full rein to the imagination of designers. Bina Ramani did her first show with cow heads on sequined dresses. Fashion has become a spectacle. It was playfulness. It was frivolous. However, in order to qualify as an industry, it must be more than a culture club.

When Valaya started, menswear stagnated to bandh-galas, sherwani and churidar. There is steady progress towards ready-to-wear and accessories. Social media guides the youngsters. Though technology arrived in fashion design, it brought an element of homogeneity. It is necessary to change the design philosophy.

Tahilani mingled with the Kumbh crowds to get fresh ideas to shape his fashion philosophy. He observed the draping styles of the pilgrims. There were ideas for slip-ons, zip ups and trouser, sari and skirt wears. There are myriads of ways of folding the cloth. In the sameness too, people are distinctive. The villagers of Kutch region too provide a lot of ideas. Valaya travelled along the silk route to get new ideas. He brought the collections inspired by Iran, Turkey and Delhi. Valaya is now controlled and minimalist.

There should be balance between tradition and modernity in fashion. Never commit the sin of bastardizing the design. But it should be blended with the language of the day.

Tahilani says he has evolved a new grammar of fashion design. One has to make a statement by a light and minimal design. Yet, it should be classical and should also go beyond the event.

Corporate tie-ups help you scale up and stay relevant. There should be workshops in the villages. India has just scratched the surface of its vast retail market. There are segments such as pret, luxury, ready-to-wear and mass. Those inspire us to carry on.

Both Tahilani and Valaya do not use showstoppers. They are confident that there are enough takers available, on the ramp and off it for what they design.

We are fascinated by clairvoyance. However, there is another tribe of futurists who see into the future, based on the status the world is in today. And there are two types of futurists — those who draw a rosy picture of the future and those who raise an alarm.

Ray Kurzweil is a futurist who predicted in 2005 that smarter machines by 2045 will lead us to an infection point which he calls singularity. As it is Kurzweil has tracked AI since 1960s. Kurzweil has preponed the arrival of singularity — from 2045 to 2029. He attributes this to algorithmic innovations and big data.

Apart from other mundane things, he expects the medical science to extend human life by several decades.

Futurists draw their power of looking ahead from their insights into looking back. To several futurists, life will be brighter ahead than ever before. And the advent of singularity will make life exponentially better.

However, all this ignores the several other trends on the horizon.

There are issues of climate change, economic inequality, polarization, misinformation. There is rise of big tech feudalism. We can ignore these issues at the cost of deterioration of life.

Kurzweil talks of nanobots in medicine which will convey drugs, genes and other payloads in the body. He visualizes a brain that will interact with the web automatically. Are we ready to ingest such nanobots? Can we conceive of computer chips inside our bodies?

There is a limit to understanding of futurists.