Blog

A deepfake is a video (or audio recording) that has been edited using an algorithm to replace the person in the original with someone else. Deepfakes are not real. It is an example of manipulation with the help of AI technologies.

Some technologies used to create deepfakes are ML, deep learning and neural networks.

Recently the deepfake video of actress Rashmika Mandana surfaced over social media platforms. Pending a new law, Digital India Act, the government asked social media firms to take down content related to deepfakes and similar misinformation from their platforms, even in the absence of a formal complaint under the provisions of already existing Information Technology Act. The failure to do so will attract Section 66D of the IT Act, 2020 (punishment for cheating by personation using computer resources and imprisonment of up to 3 years and a fine of up to 1 lac.)

Deepfakes can upend society. There are instances of fake political propaganda, news and pornography. This happened when there was no AI to assist.

Deepfakes may be derived from ‘deep learning’ and ‘fake’. It is synthetic media — media that is manipulated or wholly generated by AI.

With the advent of AI, there are many apps now accessible which facilitate the creation of deepfakes. Even there is professional assistance available at meagre cost.

Social media and smart phones spread the deepfakes at lightning speed.

Most of deepfake produced is non-consensual porn or image-based abuse.

Deepfakes affect the elections. There are videos of candidates saying something unpalatable in the final moments of an election — having the potential to change the election.

Deepfakes can cause instability in the business sector.

Disinformation campaigns can threaten the democratic processes.

In a world of deepfakes, it is difficult to distinguish between reality and fakes. It can lead to ‘infocalypse’. Public can doubt the authenticity of real media. The line between real and fake becomes hazy. We the common people will not take anything on social media at its face value. That is infocalypse — the greatest danger posed by AI and deepfakes.

In 2024, Google is going to launch a new LLM called Gemini. The launch has been announced at Google I/O Developer Conference, in May 2023. The LLM will be a joint project of Google DeepMind team which combines the resources of Google Brain and DeepMind.

It will be a more powerful LLM having capabilities to generate text, code and other creative content, translation of languages, question answering in a comprehensive way, generating poems, code, scripts, musical pieces, email, letters etc.

Gemini wants to leverage training techniques borrowed from AlphaGo including reinforcement learning and tree search. Thus Gemini will be more efficient as well as effective,

Google’s Gemini is multi-modal model that goes beyond text. It integrates several aspects of LLMs which are algorithms of deep learning and huge datasets based on natural language. And they generate new content.

Google announced Gemini in May, 2023. Gemini will be helpful to API integrators.

Google’s Gemini could compete with ChatGPT supported by the rival Microsoft. Gemini’s multi-modal capabilities will be its USP. Even input could be multi-modal. This will turbocharge generative AI. It will result into superior model performance.

Gemini may get an edge on the existing models by using tree traversal and reinforcement Learning (RL).

Gemini could also use GANs or generative adversial networks consisting of a pair of competing neural networks. They are generative as well as discriminatory.

In fact, ChatGPT and Bard are on par. Bard has been integrated with Google Workspace. It is essentially a text-based AI system. Gemini differs by being multi-modal.

Inflection went public in March, 2022. It was founded by Reid Hoffman, LinkedIn Founder, Mustafa Suleyman, co-founder, DeepMind and Karen Simonyan, a former DeepMind researcher.

As a startup Infletion AI has drawn funds from many stalwarts and companies.

Inflection-2 is a language model. It outperforms competitors such as PaLM-2 and Claude 2 and is a shade less than GPT-4. The current version is more powerful than Inflection1. Inflection1 can be compared to GPT-3.5 and PaLM-540B. Inflection 2 will soon catch up with GPT-4.

Inflection 2’s training was done on H100 GPUs of Nvidia. About 5000 chips were used. As compared to GPT-4, Inflection 2 scored 890 on the HellaSwag 10-shot, thus nearing GPT-4 score of 95.3. Inflection 2 outperforms Claude2 with chain-of-thought reasoning. Inflection 2 falls short of GPT-4 for coding and math tasks.

Inflection2 will soon run the company’s Pi chatbot.

In 2015, OpenAI was set up as no-profit company with an objective of building safe AI. In 2018, it became a ‘capped profit’ company called OpenAI LP so as to bring compute power and capital backing. There was imbalance in this formation — a non profit Board was set up to keep the company on track — keep benefit of mankind above the commercial motives.

Sam Altman was a successful CEO of OpenAI. He was terminated by this Board. Surprisingly, he comes back in just five days as CEO again. On his return, he becomes more powerful.

The reasoning for his termination-not so candid termination — not so candid communication with the Board, disregard for the safety issues or some internal politics did not sound sensible enough to keep him out.

There are other founder CEOs — Facebook’s Zuckerberg — who holds a substantial stake in the company. However, Altman is not a shareholder in OpenAI. He has won on the basis of his charisma and smartness. The manpower stood firmly behind him — X witnessed a wave of emojis through the weekend to express the solidarity of the staff of OpenAI with Altman. It also expressed the outrage of the manpower.

Any organization is what its people are, and the manpower felt strongly that OpenAI is nothing without its people. They wrote a letter addressed to the Board asking for Altman’s reinstatement, or else they would quit.

Microsoft extended support to the ousted CEO and president at lightning speed. In the process, Microsoft shares shot upwards.

Altman-like CEOs come ‘once in a generation’, and they exhibit a missionary spirit rather than a mercenary one. He is a good communicator at the same time. He is incredibly curious about things. He has set a higher objective before the manpower — achieving AGI. He provides the suitable environment in the organization to goad the manpower towards this objective. They do their work diligently. They are there for a mission.

The launch of ChatGPT in late November, 2022 was transformative step. The manpower now can ‘do and die’ for him. They are all praise for him.

OpenAI has been valued at almost $90 billion, and employees cannot be denied the big cash their stock option could bring them. Can such a lucrative deal in future be frustrated? The manpower in fact was acting in its own interest.

The Board that outsted Altman has been abolished by and large. Ilaya, the chief scientist, was full of regret for being a party to the episode. Ilaya was into research. His difference with Altman pertained to the direction the company is taking — ethical AI should be a priority rather than crass commercialization. As it is AI as a technology is coming of age, and it is working. This could have driven the wedge between the two.

On his return, Altman expressed his love for OpenAI and OpenAI too reciprocated loud and clear.

Mathematics is extensively used computer programing. The mathematical concepts make you a better programmer.

Computers use binary numeral system where there are two symbols — 0 and 1. Hexadecimal system (base-16) and base-64 are essential in programming, and understanding these is facilitated by our understanding of the binary system.

Floating point numbers are not precise, resulting into tiny errors in calculations. These numbers use scientific notations.

Logarithmic functions are used in algorithms such as binary research.

Set theory is important for working with databases.

Boolean algebra uses three operators to work with Boolean values — AND, OR and NOT.

Combinatorics facilitates the calculation all possible permulations and combinations. It is a valuable skill in programming.

Graph theory is used for network routing and optimizing various scenarios.

Complexity theory or Big O notation make us aware about the efficiency of algorithms — time and memory complexity of algorithms.

Statistics is used in ML and AI. Measures of central tendency makes you understand predictions.

Linear algebra is useful in computer graphics, neural networks. It makes us understand scalars, vectors and matrices. These are used to represent data. In cryptography, 3D graphics and ML, you have to use linear algebra.

As we know, pay TV has two major segments — cable TV, where a cable operator takes a connection from a wholesale signal distributor firm or MSM and takes the signal to individual TV sets at homes through cable. Another pay TV is DTH or direct to home through a satellite dish. The total pay TV market was 160 million homes in 2020, which has now reduced to 100 million homes.

There is leakage at both the ends. Viewers are going to streaming platforms through broadband internet (OTT channels). These channels are offered by Jio and Airtel. At the other end, there is free DTH of Prasar Bharati through DD Freedish.

It means TV watching is done more through streaming platforms received through broadband connection or through freedish. The DTH or cable connection suffer in the process.

You should have a smart TV connected to broadband to watch both OTT channels and regular TV.

Subscriptions of DTH channels are falling down. To counter this, Tata Play in 2020 launched Binge offering Netflix and 20 other OTT channels in bundles. They provide a special set-top box for this . Airtel Xstream also provides 18 OTT channels.

Aggregation of OTT services is a natural transition to improve average revenue per user.

There are wired broadband homes in India. Many of them have smart TVs. A bulk of viewing happens on smart phones. These consumers do not bother about smart TVs and wired broadband.

TV screens ad rates are four to six times than those of the mobile screens. TV screen could be of ordinary linear TV or smart TV. Tata Play used Tata Broadband to operate in the last mile connectivity to reap the benefits, but the big players have realised that fibre laying costs are very high, and the payback period is long. That is the reason why cable operators are used for last mile pipe. Still in bigger cities operators do provide last mile connectivity, as the quality of fibre is good.

DTH industry has stabilised and would recover, but it may not have double digit growth.

AI requires careful handling to benefit humanity. At its worst, it can become an existential threat for humanity. AI can facilitate autonomous weapons systems which can identify targets with precision. AI can generate surveillance systems to help authoritarian regimes. AI can enable cloning abilities among criminals to carry out scams.

At the same time, AI can do a lot of good. It has the potential to be a game changer. It can perform hard tasks such drug discovery, management of nuclear power plants and telecom networks, management of power grids and road and air traffic, management of mining.

In a couple of years, NASSCOM puts the benefit to India due to AI-related activity at $450-500 billion to India’s output.

Thus, both the boardrooms and regulators will have to make numerous adjustments while dealing with AI.

The employees of OpenAI were overwhelmingly in favour of the outsted CEO, since he was seen to be driving the commercialization of ChatGPT which is not much ahead of Bard, Llama or Grok of competitors. The employees want to encash their skills, and get stock options if the subsidiary gets listed. It is normal to expect this in Silicon Valley.

Altman, after his return from a few days exile, had a hero’s return. Before the turmoil, the six member Board had three staff members and three independent directors, and was commited to effective altruism. After the turmoil, the Board with nine members may have at least one seat (say as an observer) for Microsoft. It is not yet clear how the independent Board members responsible for Altman’s ouster will be dealt with. They have agreed to step down. Though Quora’s D’Angelo voted against Altman, he continues after change of heart. OpenAI’s structure as a non-profit organisation and a for-profit company is difficult to balance. And the Board answered to the moral instincts only. This set-up was problematic for Microsoft, with 49% stake in OpenAI with no Board seat. The CEO will be more accountable to Microsoft henceforth. The new Board will neutralise the overarching influence of the Board on for-profit activities of the company.

Microsoft’s creating an OpenAI team would not have worked so well. As an independent unit OpenAI so far takes all the flak for corporate reputation and legality while deploying ChatGPT and DALL-E2 in the market. OpenAI remains a startup, and can fiddle with cutting-edge AI technology.

Microsoft keeps intact all its glow. and with none of its liability. Besides, it can exercise more control. It is a better deal for Microsoft to have OpenAI as a free unit than to have OpenAI inhouse.

A story has been floating around that some employees wrote a letter to the Board pointing out the development of an algorithm that could be a breakthrough in AGI. It solved certain math problems. It makes them optimistic about future success. It solves the problems for which it has not been trained. It means it has certain reasoning abilities. The project was code-named Q* or Q Star. In fact, the company was pushing the veil of ignorance back and and the frontier of discovery forward. However, this triggered Altman’s ouster.

There are unfair trade practices in travel industry. They are the patterns that lure consumers to opt for something that involves additional expenditure. Travel websites and airlines can show all the seats in a flight ‘paid’ when passengers are asked to do online checking before the travel. The design interface is so made. The consumers do not know ‘paid’ seats are not mandatory and there could be ‘free’ seats. It makes a consumer opt for a ‘paid’ seat and pay the seat allocation fees.

In addition, travel websites do not disclose that booking through them attract ‘convenience fee’.

There are complaints that though tickets are cancelled but refunds do not come from the airlines. There is delayed compensation for the lost or damaged luggage. There are unfair charges for the aisle or window seats. There is the dark practice of forcing the passengers to insure their trip.

It is necessary to protect consumers from misleading or coercive online tactics. In 2010, a phrase ‘dark patterns’ was coined by Harry Bringall, a user experience (UX) expert. Dark patterns refer to user interfaces (UIs) designed intentionally. The aim is to deceive, manipulate or compel users into specific actions. These run contrary to their preferences.

Mostly manipulative practices, these leverage cognitive biases — perceived scarcity, urgency, the need for validation. Users are thus prone to act hastily. There is fear of being left out.

Dark patterns have invited attention of the authorities. The European Union’s Digital Services Act aims to ban dark patterns. The UK has started investigating such practices. The US is concerned about deceptive design. In India, there are guidelines for prevention and regulation of dark patterns.

When consumers are led to unintended actions, their autonomy is undermined.

Some illustrations of dark patterns are subscription traps, interface interference (manipulation of information presentation), bait and switch tactics, false urgency (only 25 products are left), basket sneaking ( adding items without the knowledge of the user, surreptitious addition ), forced action (imposing additional purchases), use fear or guilt to influence user actions, drip pricing (price is concealed or delayed disclosure of price).

The guidelines are issued in exercise of powers conferred by the Consumer Protection Act, 2019. These are in fact an extension of the concept of ‘unfair trade practices’.

Indian transactions are based on trust, and the dark patterns erode the trust.

Despite the guidelines, what matters is the enforcement. There should be a robust enforcement mechanism. Regulators should stay ahead of new deceptive tactics.

Nagging falls between marketing and undue interference with the user experience. It is a subjective concept. There is litigation on such areas.

Guidelines align with the global efforts to fight deceptive design practices.

OpenAI’s ouster of Sam Altman at the instance of a non-profit Board on Friday had the unpredictable effect on the AI talents the company had — they threatened to resign en masse if the Board refuses to reinstate him.

Ilaya Sutskever, the chief scientist played a bizarre role and he is the one who asked Sam to quit. Later he came into a self-correcting mode and tweeted his intention is ‘not to harm OpenAI.’

In the meanwhile, Microsoft was ready to accommodate the oustees and appointed Sam as the CEO of a new AI structure.

As we know, OpenAI being valued $90billion, has been thrown into a turmoil by its overarching non-profit Board. It seems Sutskever loves being esoteric, scared of AI attaining the god-like status. He would not like to have an unaligned AI contrary to the interests of humanity.

It is a strange case of a company where the state-of-the art technology is being hampered by some untested ideas.

The situation is pretty chaotic. Maybe, the Board imagined that there is rapid hurtling towards singularity. Could it be a locker-room mentality where rival views are expressed by colleagues?

It is a story with many twists and turns.

It is now learnt that the new Board of OpenAI has reinstated both Sam Altman and Greg Brockman in view of the supporting stand taken by the majority of existing employees. The new initial Board has Bret Taylor, Chair, Larry Summers and Adam D’Angelo. Bret Taylor is former, Salesforce co-CEO.Larry Summers, is former US Treasury Secretary. Adam is from Quora.

Though the technology AI is likely to flourish, this saga at OpenAI puts doubts in our minds about the future of OpenAI.